← Back to Archives
VOLUME 15, ISSUE 10, OCTOBER 2026
Publication in progress..
VALUDRIVE - VEHICLE PRICE PREDICTION AND RECOMMENDATION SYSTEM USING MACHINE LEARNING
K Suraj, Shruthilayaa M, Vidya Sree, Dr. Beenarani Manoj
Heart Disease Prediction Using Machine Learning with Explainable AI
Karthi Balaji M, Nithila K, Sandhya Dillybabu, Dr. Beenarani Manoj
A Survey on SLM for stylized Marathi Prose and Poetry Generation
Manasi Bhole, Riddhi Lahare, Veda Kimbahune, Vedika Kute, P. P. Joshi
GEOVALUE PREDICTOR HOUSE PRICE PREDICTION FOR INDIA
Dr. Beenarani Manoj, Thanish L.S., Sushaanth.M, Durai Raj.R, Tharun V S
Tamil Named Entity Recognition Using Classical Machine Learning, Quantum Feature Processing, and a Simplified QNLP-Inspired Component
Fexi Maxwell G, Paavai Anand G, Sridevi S
Intelligent Anomaly Detection for Digital Transactions Using Machine Learning
Desale Harshvardhan, Dr Pradip Patil, Prof Kiran Chaudhary, Namita Devare, Keshar Patil, Divya Jagtap
Design and Performance Evaluation of a Post-Quantum Password Manager Using Argon2id, AES-256-GCM, and ML-KEM
Sanjana Pawar, Rama Pandey, Mrs. Soly Zacharish
Green IT: Reducing the Energy Use and Carbon Footprint of Institutional Computing Infrastructure – A Case Study of a Mid-Size Campus Data Centre
Sahil Patil, Sahil Mohite, Dr. Nisha Yadav
Detection of Glaucoma using Machine Learning Approach
Zaki Hasan, Prof. Syed Akhtar Hussain, Zafar Ul Hasan
A Review of Deep Learning for Real-Time Sign Language Recognition
Minal V. Patil, Pravin I. Patil, Dr. D. M. Patel
A Novel Hybrid Dynamic Pheromone Optimization Algorithm with Entropy-Guided Learning and Historical Route Intelligence for the Travelling Salesman Problem
Mrs. M. Hemalatha, Dr. N. Kamaraj
Neocube Realty: A Web-based approach for comprehensive real estate Business System
Pranjal Rane, Muskan Madhwani, Vaishnavi Sawarkar, Kajal Gawande, Chanchal Karwa, Achal Pande, Prof. S.A. Solanke
Fake News Detection Using Machine Learning and Natural Language Processing
Dhanashri Satish Deshmukh, Ms Rucha Mohite
Bridging the Language Gap: NLP Based Translation of Government Schemes for Marathi- Speaking Farmers
Pooja Suryabhan Gaygaye, Ms.Yogita Kadbane
SIX-YEAR ENROLLMENT TREND ANALYTICS AT WESLEYAN UNIVERSITY-PHILIPPINES: A DATA-DRIVEN BASIS FOR REGISTRAR'S OFFICE DIGITALIZATION AND AUTOMATION
Yolanda M. Javier, Shirley DC. Ramos, Abigail C. Bautista, Carmeliza S. Valdez, Rozen Jake DR. Punay, Gemmuel M. Borja, John Andrei J. Ignacio
Automated Multi-Student Attendance Using Video-Based Facial Recognition and Embedding Matching
Likhitha Alajangi, Dr. Kuda Nageswara Rao
AI-Based Semantic Understanding of Code Mixed Tamil-English Text in Informal Digital Communication
Aditya Raman, Sushmita Kundu, Aditya Balaji, G.Paavai Anand
Comparative Analysis of Convolutional Neural Network Architectures for Automated Waste Classification
Sakshi Anil Parkhe
MACHINE LEARNING FRAMEWORK FOR DETECTION OF MENTAL STRESS
Archana Borse, MS. Yogita Kadbane
HemaSight AI: A Four-Stage Deep Learning Pipeline for Automated Leukemia Detection and Post-Transplant Monitoring
Varun Doddagoudar, Vishwa, Shashank, Madhushree M
Smart Irrigation System for Precision Farming
Mr. Bhagyesh N Chaudhari, Dr. Dinesh D Puri
Multi-Crop Disease Detection and Management Recommendation System Using Deep Learning
K. Hemalatha, Prof. K. Venkata Rao, P. Swathi
CUSTOMER COMPLAINT TRACKING SYSTEM
Dr. L. Baby Victoria, Abinaya D
Crop Recommendation System for Farmers Using Convolutional Neural Network
D. Savitha, Gayathri U, Priya S
Equilibrium Dynamics and Threshold Conditions in Tumor–Immune Models with Chemotherapy
Shivangi Chauhan*, Prof. Diwari Lal
Quantum Assisted Predictive Maintenance and Scheduling System for Industry 4.0
Muppuri Satish Kumar
Chronic Kidney Disease Prediction System Using Machine Learning
Savitha. D, Haritha. V P, Mathisha. N
Women Enterprise Hub: A Web-Based Platform for Women Entrepreneurs
D. Savitha, S. Nitharsana, S. Subarna
DESIGN AND DEVELOPMENT OF AN
ESP-BASED SMART MOBILE, MANIPULATOR FOR REMOTE OBJECT, HANDLING, Kuncham Preetham, Dr. K. Raja Kumar
An Intelligent Risk-Adaptive Framework for Secure Examination Paper Distribution, Traceability, and Early Leak Detection
Mr. Swapnil S. Chaudhari, Asst. Prof. Sapana A. Fegade
Smart Classroom Seat Allocation System Using IoT, Artificial Intelligence, Machine Learning and Optimization Techniques: A Review
Dr. P. Anitha, V Karishni
PERSONALISED HEALTHCARE RECOMMENDATION SYSTEM USING XGBOOST AND SHAP
Savitha. D, Sowmiya. V, Akshaya. C
Real-Time Sign Language Translator Using Video Prediction
Geetha. N, Mohanashankare.V.K, Prathanyaa.K
Agent-less Windows System Vulnerability and Network Scanner for Security Assessment and Reporting
Satvik Dilip Barhate, Dr. Dinesh Puri
Healthcare Insurance Fraud Detection Using Machine Learning and Deep Learning
Siddanathi Prasanth Srinivas Sai
Grouping-Aware Numerical Verification for Offline Retrieval-Augmented Question Answering over Institutional Documents
Sudheer Katuri, G. Sharmila Sujatha
Genuine Co-Movement, Spurious Trend, and a Ratio That Does Not Help: Variable Construction for Nigerian Macroeconomic Models
Ofierohor Ufuoma Earnest
Abhimanyu Chakravyuha: An Evidence-Aware AI Framework for Fraud Intelligence Through Semantic Analysis, Mutation Detection, and Graph-Based Reasoning
MD Auranzeb Khan, Kunal, Srijan Mani Tripathi, Vyom Pandey, Dr.G Paavai Anand
AI-DRIVEN CLINICAL DECISION SUPPORT SYSTEM FOR INTELLIGENT MEDICAL REPORT ANALYSIS AND DISEASE PREDICTION
Sravanthi Duvvuri, Dr. G. Sharmila Sujatha
AI-Enabled Structural Vibration Prediction for Enhanced Stealth Performance
Harish Kumar Kuppili, Kuda Nageswara Rao
Contract Sentinel: A Literature Survey on Transformer-Based Clause Extraction and Risk Detection in Legal Contracts
Vaishnavi Karpe, Tanishka Makode, Maria Shaikh, P. J. Jambhulkar
SR-Screen-AI: An Explainable Rule-Based AI-Assisted Decision Support System for Human-in-the-Loop Systematic Literature Review Screening
Sruthi R, Rajeev P R
How Many Normal Images Are Needed? Training-Free Few-Shot Anomaly Detection on Scanning Electron Microscope Images of Integrated Circuits
K. Susmitha Reddy, G. Lavanya Devi
A Nightshade-SWIN Transformer-Based Framework for Enhanced Multi-Crop Plant Disease Classification
Dr. Barkha Joshi
A Secure and Privacy-Aware Hybrid Framework for Digital Credential Verification
Jayashri Digambar Sable, Swaranjali Dinesh Shelar, More N.N
An AI-Based Intelligent System for Early Urban Health Risk Assessment
Mali Rajashri Savata, Salunkhe Samruddhi Mahadev, Lokare Sujata Sharad, Ghalame Mayuri Kishor, Linge Sandeep Ganpat
Self-Evolving AI Agent Framework for Predicting and Preventing Zero-Day Cyberattacks in a Multi-Cloud Environment Using Federated Learning and Explainable AI
Samruddhi Santosh Gawade, Gayatri Gunwant Gayaki, Ms. Ghodake G.K
School Management System for Dutuwewa Maha Vidyalaya
Keshika Rathakrishnan, Parahera Nekathige Kaumini Viduthala
A Voice-Powered Intelligent Healthcare Platform for Elderly Patients with Dementia and Alzheimer's Care
Shinde Akshata Kumar, Pathan Ekhara Mushtak, Parchande Sanika Harishchandra, Khilare Sakshi Kalidas, Sawant Namdev Manik
MetaSociety: A Simulation-Driven AI Policy Advisor for Data-Grounded Smart Governance
Priyanka Jadhav, Payal Dongre, Priti Chaugule, Mayuri Kulkarni
Smart AI-Powered E-Commerce Platform with Personalized Recommendation and Sales Analytics
Wadekar Gargi Prashant, Bole Ishwari Santosh, Patil Harshvardhan Sunil, Mr. S. S. Ingole
IOT-Based Machine Health Monitoring System
Chavan Rutuja Umesh, Babar Dnyaneshwari Namdev, Gajare Pragati Dnyaneshwar, Mrs. P. S. Gavali
AI-Based Road Accident Detection and Alert System
Mr. S. A. Zambare, Saniya Mainuddin Mulla, Kiran Sanjay Jadhav, Shivprasad Suresh Kadam, Reshma Mujofar Pathan
Smart Village Automation Using IoT
Ms. J. K. Hipparkar, Pooja Karande, Shreya Joshi, Arjiya Mulani
VocaLyze – AI Based Emotion Recognition from Voice
Sadigale Shravani Santosh, Mali Vaishnavi Sukhdev, Gotavle Sushma Sanjay, Linge Sandeep Ganpat
An Adaptive and Risk-Aware Access Control Framework for Secure Cloud File Sharing Using Hybrid Cryptography and Context-Aware Key Release
Ghodekar Snehal Ramesh, Ghodekar Vaishnavi Suresh, Miss. Ghodake G. K.
Information Security Copilot for Cyber Threat Monitoring and Risk Management in Small and Medium-Sized Businesses
Yashowardhan Thakare, Rutuja Ghongade, Payal Pawar, Radha Mankar, Anisha Kaware, Dr V B Gadicha
Smart Public Infrastructure Damage Reporting and Prediction System: A Web-Based Platform with Category-Based Rule Priority Classification
Khushi, Harsha Bhat T.M, Varun P, Shreecharan I G, Prof. Raksha K
AI in Healthcare Applications: A Study of Patient Awareness and Acceptance
Priyanka Ambdekar, Aditi Hingane, Prof. Pramod Gopale
CNN-Based Framework for Maize leaf Disease Detection and Crop Monitoring
Dr.K.L.S.Soujanya, V. Sumedha, M. R. V. Sravanthi, A. Supriya
A ROBUST FRAMEWORK FOR COPYWRIGHT PROTECTION AND OWNERSHIP VERIFICATION OF DIGITAL IMAGES
Pavithra N, Anushree P M, Kavana K, Megha Manjappa Mogali, V Tejaswini
An Intelligent Machine Learning Framework for Early Cardiovascular Disease Risk Prediction
Gnanamani. H, Gagana G, Inchara Bhatta K.M, Kavana, Likitha. H
DETECTING AND MITIGATING HALLUCINATIONS IN LARGE LANGUAGE MODELS: A REVIEW OF UNCERTAINTY, EVIDENCE GROUNDING, AND AGENTIC WORKFLOWS
Anindya Singh, Rishiraj Singh
Blockchain Based Academic Certificate Verification System
Rutuja Dudhate, Shubham Gadge, Pramod Gopale
AI-Based Fraud Detection in Digital Payment Systems
Puja Patil, Shruti Pingat, Prof. Nikita More
A Review of AI-Driven Cyber Emergency Response Systems: Edge Intelligence, Digital Cyber First Aid, and Human-Centric Cybersecurity
Devendra Singh Kushwaha, Uday Payghon, Manjushree Dighe, Riya Pagar, Anushka Shinde
AutoAssist AI: An Integrated Intelligent Platform for Vehicle Assistance and Automotive Marketplace Management
Vedant Virender Kalpande, Mohammad Huzaifa, Akhilesh Pimpalgaonkar, Abhishek Shende
Developing an Integrated Big Data Analytics Framework for Personalized E-Learning in Open and Distance Learning Institutions in Kenya
Antony Odhiambo Okeyo*, Daniel Khaoya, Bostley Muyembe
A Voice-First Personal Finance Assistant Using Tool Calling and a Locally Served Fine-Tuned Small Language Model
Rishi Raj Sharma, MD Faiz, S. Madan, Tejashwini, Udaya Shankar Budagavi
Demand & Decision Intelligence System: An AI/ML-Based Platform for Demand Forecasting, Anomaly Detection, and Inventory Optimization in Indian Retail
Prof. L.S. Bhattad, Yash Gawande, Vaishnavi Chauhan, Qazi Ubed Rehaman, Yash Zade, Kaushal Tayde, Disha Patil
VitaDetect: Deep Learning Based Vitamin Deficiency Detection System
Anchith Shanthivana C, Anjali V, Lakshmi Arpan, Pratham B G Rai
A Reinforcement Learning-Based Approach to Optimize Network Performance in Wireless LANs
Dr VijayaLakshmi D, Manju L, Manya V Bethur, Niriksha V Shetty, Pragathi P
Multimodal Industrial Predictive Maintenance Using Hybrid Deep Learning and Explainable AI: A Literature Survey
Abhijith Nambiar, Ajin Ani Philip, Akshay Mahesh, Kevin T John, Merlin K Thomas, Girija V R
AutoSecAudit: An Extensible Multi-Scanner Orchestration Framework with AI-Assisted Remediation Guidance, Precision Fuzzing, and Delta Auditing in DevSecOps
Kodi Vishruth Aithal, Thilak N, Vinaya Kumar B, Vishal Gowda B, Ranjitha R*
SentinelML: Behavioral Machine-Learning-Based Ransomware Detection with Decoupled AI Security Analysis
Vasu Chowdary, Moulica Chowdary, D. M. Yashaswini, Rajashree M Byalal
MAHATVAK CHIKITSA- AUTOMATED SKIN ONCOLOGY DIAGNOSIS WITH CLINICAL DECISION SUPPORT
Dhanya NG, Manoj R, Megha R, Harshitha R
AI-Powered Smart Glasses for Digital Eye Strain and Drowsiness Detection
Sanmaya I K, Sharanya S, Vinusha D Souza, Maneesha, Bharathi K
Comparative Analysis of Deep Learning Models for Offensive Comment Detection in Code-Mixed Dravidian Languages, with a Focus on Tamil-English
Maria Sarah J, Dr. Golda Dilip
Cyber Crime Awareness, And Fraud Prevention
Nilam Yashwant Kokanee, Asmita Bhaskar Kudnar, Prof. Sanchita Panmand
Evaluating Privacy–Utility Trade-offs in Differentially Private Federated Learning: A Comparative Benchmarking Study
Divya Varshini M, Dr Beenarani Manoj
Carbon FootPrint Tracking in Logistics System: AI-Based Logistics Management Platform
Bhagyashri K Kulkarni, Shashank Urs H P, Vinod, Sandeep T Rathod, Mayur Kishan Rathod
Phishing Website Detection using Machine Learning
Prof. More N.N., Wayal Akanksha Sandip, Walunj Sneha Bhausaheb
Explainable AI-Based Early Dropout Risk Detection and Intervention System for Rural Schools
Pranali Santosh Wagh, Disha Santosh Waghmare, Miss. More N. N.
The Price of Credit Risk in Nigeria: Do Bank Lending Rates Reflect the Opportunity Cost of Treasury Investment? Evidence from Lending–Treasury Spreads, 2006-2026
Ofierohor Ufuoma Earnest
Context-Aware Conversational AI for Intelligent Travel Planning and Smart Tourism
Arathi H L, Aishwarya A U, Deekshitha H R, Harshitha S, Mahesh Hanchanl
Hybrid Random Forest–Marine Predator Algorithm for High-Accuracy Energy Theft Detection
Devaraj V, Akila K
A Comparative Analysis of Prediction-Based and Reconstruction-Based Deep Learning Models for Time Series Anomaly Detection in Spacecraft Telemetry
Ashwin Krishna N, Dr.H.Mary Shyni
Beyond Contextual Embeddings: A Comparative Analysis of Feature-Engineered Classical Machine Learning and Transformer-Based Approaches for Deceptive Opinion Spam Detection
Shaliny P, Dr.S.Niveditha
Intelligent Edge AI System for Satellite Image Analysis and Adaptive Data Transmission
Hari Krishna M.B, Siddarth.A, Niranjana G.M, Santhosh, Udai Shankar
AI-driven Renewable Energy Generation Forecasting Using Machine Learning Techniques
Arathi H L*, Sinchana C S, Partha C M, Shobha N, Shreya K S
Drift-Aware Online Ensemble Learning for Imbalanced Industrial Failure Detection: A Comparative Study
Janani M, Dr.M.Daniel Naraeshkumar
A Comparative Study of Traditional and Ensemble Classifiers for Software Defect Prediction on the NASA PROMISE Repository
Sanjay I, Dr.P.Durga Devi
RSAA-UNet: A Residual Squeeze-and-Excitation Attention U-Net with Atrous Context and Focal-Tversky Supervision for Lower-Grade Glioma Segmentation on FLAIR MRI under a Patient-Wise Evaluation Protocol
G. Appasami*, A. Selvakumar
AI-Powered Image Captioning and Content Generation: Design, Implementation, and Evaluation
Hirek Praful Warhekar, Kunal Dinesh Afare, Manasvi Dharmendra Ingale, Sagar Sanjay Kinada, Anup G. Kadu
Design and Validation of a Cognitive 360° PLM Framework for Autonomous and Adaptive Product Lifecycle Decisions
Rahul Chandrayan
Braive: IoT Enabled Smart Braille Learning Glove for Assistive Education
Swathi, Thanmayi Rai A, Thashvi Rai K, Vibha H D, Deepthi M B
Artificial Intelligence and its Applications
Sakshi Gorakh Aher, Jitesh Bhimrao Ade, Prof. Gopale. P. J
Machine Learning and Explainable AI Techniques for Smart Grid Stability, Security, and Decentralized Energy Management: A Survey
Basima S, Hanna Fathima H, Ninu Bijoy, Sreemayi Saju, Sravana Pradeep
Low-Cost Wi-Fi CSI-Based Human Activity Recognition Using ESP32 and a Gated Recurrent Unit Classifier
Pavani M B, Priya M S, Vindhya G M, Jeevan H G
CareerCraft AI: An Intelligent Employment Support System Integrating Resume Building, Applicant Tracking System (ATS) Analysis, and GeoBased Job Matching
Samidha S. Shende, Srushti S. Khate, Srushti M. Deshmukh, Anuj G. Gondane, Prof. N. K. Chede
The Role of Artificial Intelligence in Digital Marketing
Akshada Gorde, Aditya Hilal, Prof Ghodke. G. K
SmartCompress AI: An Entropy-Aware Hybrid Machine Learning Framework for Intelligent Lossless Compression Algorithm Selection
Sanjiv R, Sriramji P, P. Durgadevi
Application of Computer Vision Technology in Industrial Automation
Aishwarya Solanke, Bhumika Solanke, Prof. More. N.N
An AI Powered Automated Code Review System Using Large Language Models and Static Analysis
Mrs. Nita Meshram, C. Balaji Naidu, Ganni Naveen Raj Anudeep, K. Vedavyas, K. Harsha Nandhan
LLM Red-Teaming and Evaluation Agent System
Chanchal S. Pophale, Mayank A. Darokar, Prashansa P. Patil, Prathmesh S. Tayade, Neha Kandalkar
Artificial Intelligence and Robotics for Automated Pharmacy and Medicine Dispensing Systems: A Survey
Abhijith T S, D H Aromal, S Nasif, Nithul Das H, Neethu S, Dr.Shani Raj
AI Based Interview and Evaluation System
S Asif Hussain, Mizba, Mohammad Kouser, Mohammed Mehraj Fathima, Vijay Kumar R
MediCycle AI: A Smart Framework for Safe and Sustainable Redistribution of Surplus Medicines
Mrs.Anjana H S, Atharva K R, Gnanasagar G N, Jeevan P, Siddeshwar B V
A Survey on Federated Learning for Privacy-Preserving Healthcare: Towards Differentially Private Early Sepsis Prediction
Harshil P, Sreehari R, Kesiya Shaji, Adivinayak M V, Sukanya M V, Girija V R
Machine Learning-Based Procurement Price Benchmarking Through Historical Ranking and Market Evidence
Keshava A, Amrutha N V, Anu Patel M V, Deepthi M, Shashank S
SentinelGuard: Explainable Detection of Spyware and Remote Access Trojans on Windows
Nirbhay A. Shende*, Vishal V. Ghorad, Soumya P. Mendse, Kanak V. Gupta, Radhika P. Malu
A Hybrid AI Recommender System for Dynamic Career Guidance
Addhwaith S Ajith, Aaron John Joy, Vaishnav Biju, Ameesha J S, Jency Mariam John, Girija VR
From Fragmented Capability to Unified Investigation: A Critical Review of Artificial Intelligence Across the Police Investigation Lifecycle
Goury S, Anjana V, Nythal A, Roobini R, Girija V R, Rini V R
Aegis: An AI-Based Network Intrusion Detection and Prevention System with Stateful Scan Detection and ISO/IEC 27001 Incident Reporting
M. Sreedharan, Lokesh S, Deepashri P, P. Durgadevi
A Comparative Study of LWE-Based Post-Quantum Key Encapsulation and Traditional RSA
Rutuja Paraji Zaware, Miss. More N. N.
AI-DRIVEN RESUME ANALYSIS AND PLACEMENT SUPPORT
Anurag AP, Uthara Nair, Sreelekshmi, Amanda Terence, Siji Chacko, Girija VR
RansomGuard AI: An AI-Powered Ransomware Detection and Storage Protection System
Rithvik Rajmohan, Roheith R, Bharath R, P. Durgadevi
From External Grounding to Collective Intelligence: A Comprehensive Survey on Factuality, Knowledge Graphs, and Multi-Agent Orchestration in Large Language Models
Abishek S, Alfred Jaison, Ashish S, Karthik B, Anpu Alexander, Jinu Lenin
AI-Based Smart Classroom Attendance System using Face Recognition with CCTV Camera
Praveen Kallappa Ullagaddi, Rahul K G, Vikhyath, Vishal Paramavva Madar, Prof. Pradeep Kumar K G
MediGuard: A Safety-First AI-Assisted Healthcare Companion Platform
Vidhi Ingle, Pratik Wankhede, Sayali Deshmukh, Sujal Jadhav, Prof. R. C. Awatade
DESIGN AND PRELIMINARY FUNCTIONAL EVALUATION OF GUARDIAN ATTENDANCE: A BROWSER-BASED FACE RECOGNITION SYSTEM FOR EXAMINATION-HALL ATTENDANCE
Dhanush S, Nirmala Basavaraj Mali, Preethi V M, Rahul N, Nikhil G
ROMEO: Raw Onboard Multispectral Earth Observation — An Edge-AI Framework for Latency- and Bandwidth-Efficient Satellite Sensing
Charan M H, Abhishek G, Haridarshan K P, Mallikarjuna, K B Janaki*
AI Powered Sign Language to Text and Speech
Samanvi I, Sanjana Hegde, Shravya K V, Shravya N Shetty, Bharathi K
SasyaVedha: A Soil-Aware and Climate-Resilient Agronomic Decision Support System for Indian Farms
Nishanth U, Lalith N, Punith Y D
Agentic Commerce: A Survey of Emerging Landscape
Janhavi Manore*, Sharvari Jamkar, Nupoor Joshi, Sukanya Gupta, Parag Jambhulkar
A Survey on Privacy-Preserving Face Recognition: Techniques, Challenges and Future Directions
Gopika S Nair, Sebin Sibichan, Sri Sai G Shyam, Stanly Antony V W, Sumi Rehim, Dr. Shani Raj
AutoFix: A Verification-Driven Framework for Automated Program Repair and Security Patching of Python Code
Pranav K, Devabala Kurup S, Rekha P S, Abel Markose, Sidharth S, Soja S Padmam, Girija V R
Blockchain Based digital forensics for Evidence Collection and Provenance Preservation
Vaishnavi Shankar Kalamkar, Sanika Dattatraya Kalokhe, Miss. Ghodke. G. K
Sign-Speak: Real-Time Continuous Indian Sign Language Translator
Rashmi N D, Yashaswini H L, Sinchana M S, Madhushree M, Lavanya N L
An AI-Powered Lie Detection System Using Facial Micro-Expressions
Mrs. Recharla Surendra Geethanjali, Aditya Prakash Masabinal, Bharath Kumar S C, C Yuvaraj, G Daewoo Sri Prasad
AI DRIVEN SUSPICIOUS ACTIVITY RECOGNITION USING DEEP LEARNING
Neelakantappa T T, Meghana G S, Nandini G, Gousiya Fathima A, Anusha R H
UrbanPulse: A Dynamic Multi-Attribute A* Pathfinding and Real-Time Telemetry Framework for Intelligent Urban Bike Navigation
Arathi H L, Varun P M, S Rahul, Shashank P R, Omkar S
RESQMESH: AN OFFLINE EMERGENCY COORDINATION SYSTEM
Pushpalatha M, Pavitra N, Ramyashree S S, Nithya S, Sirisha D, Varshini R
A Survey on IoT and Artificial Intelligence-Based Smart Home Energy Management Systems: Techniques, Challenges and Future Directions
Athul Krishna R, Darvesh Muhammed N, Naveen P Jose, Sujith Balachandran, Aswathy S, Dr. Shani Raj
AI-Powered Automated Code Review for GitHub: A Survey of Static Analysis, Large Language Models, Hybrid Approaches, and Developer Workflows
Archana Raj, Pavithra J, Hima Shaji, Jyolsna K V
CareBridge – Unified Healthcare Services System
Mr. Anand Rajendra Katariya, Miss. Shreya Gajanan Lawange, Miss. Sania Vivek Raut, Mr. Swaraj Prashant Chandurkar, Miss. Priya Rajesh Bawrani, Dr. Gaurav J. Sawale
Early Detection of Alzheimer’s Disease with Blood Plasma Proteins Using Support Vector Machines
Sharanabasavaraddi B, Spandana KP, Yashaswini LM, Madhuri JK
ThreePod: A Verification-First System for Prompt- and Image-Driven 3D Modelling and Printing
Sharvil R. Wange*, Lokesh A. Chaudhary, Jay P. Umbarkar, Hemant N. Humane, Shailesh S. Dhok
AgriCopilot: An Offline-Capable, Edge-Native Single-Page Architecture for Precision Agronomy and Inter-Market Price Routing
Punitha M R, T Ranadeep, V Bhuvan Chandu, Shreyas R V, Yuvan B
KubeVision AI: Kubernetes-Native Observability and Root Cause Analysis Platform
Shubham More, Prajwal Mandlecha, Kaustubh Gharat, Kaushal Waykole, Prof. P.A.Jain
AI-Powered Penetration Testing and Vulnerability Remediation for GitHub Repositories
Mrs. Bindu K.P, Gagana R, Kushal K R, M G Sahana, and M Harshit Pramod
Measuring Web Tracking and Browser Fingerprinting Across the Popularity Spectrum: A Stratified Crawl of 20,911 Websites from the Tranco Top Million
Krushna Sanjay Mohod, Atharva Sunil Raut, Atharv Rajesh Kakade, Shravani Sachin Yawale, S. R. Gupta
RankSetu AI: AI-Powered Smart Learning, Practice & Exam Readiness Platform
Dr. Suryavanshi Arti P., Vaishnav Gorakh Funde, Shahista Irfan Tamboli, Gaurav Balbhim Jawale, Aditya Raju Jadhav
NeoLipi: A Multi-Script Ancient Inscription Recognition and Translation System Using Deep Learning
Ramya Vinayak Naik, Sahana R B, Sakshi Sanjay Hegde, Sathwika M D, Prof. Shrinidhi A
STOCK PRICE PREDICTION
Shree Rakshana R, P. Aarti Pai, Akrithi RJ, Dr. Beenarani Manoj
REVIEW PAPER ON APLICATIONS OF GRAPH NEURAL NETWORKS AND BIOMEDICAL LINK PREDICTION
Atharva Prafulla Pagey, Diya Baldota, Rajlaxmi Take
Unmasking Deepfakes: Audio-Visual Deepfake Detection Using AI
Akhila P. V, Sahana J, Sagarika Sujil, and Sreehitha P. C
AI-BASED SKIN LESION CLASSIFICATION AND REPORTING USING DEEP LEARNING
SharanaBasavaRaddi B, Nirmitha P, Sushmitha H, Chinmai B, Thara R
Downfeather: A Machine-Learning-Driven Adaptive Multi-Connection Download Manager
Jatin Verma, Abhishek Jatla, Thanuja N BS, Dr. Beenarani Manoj
AI-Driven Smart Agriculture Platform for Integrated Crop Health Monitoring and Yield Optimization: A Systematic Literature Review
Rakshith Prabhu, Srihari M.R, Srinivas Gowda C.N, Suhas C, Prof. Anitha L
Trading Copilot for Cryptocurrency Markets: A Review of Analytics, Prediction, Manipulation Detection, Automated Execution and Conversational Agents, with a Gap Analysis of Existing Systems
Sanskruti Patil, Mokshada Sheth, Shraddha Kokare, Naveli Pangal, Kopal Gangrade
Abstract
VALUDRIVE - VEHICLE PRICE PREDICTION AND RECOMMENDATION SYSTEM USING MACHINE LEARNING
K Suraj, Shruthilayaa M, Vidya Sree, Dr. Beenarani Manoj
Abstract: Used-car resale prices are veering according to vehicle attributes such as year, kilometres driven, fuel type, transmission, engine capacity and ownership history. This paper introduces ValuDrive, a machine-learning web application that assessments used-car selling prices and recommends comparable vehicles in one interface, to a important degree, in many respects. Linear Regression, Decision Tree Regressor and Random Forest Regressor were inspected employing an 80:20 training, testing split and the metrics Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R². This model is using the Car details v3.csv dataset and applies data cleaning, feature preparation, numerical imputation, categorical imputation and One-Hot Encoding, in real-world scenarios. Random Forest hit the strongest findings among the found models, with an MAE of ₹80,195, an RMSE of ₹180,544 and an R² of 0.851. The application blends the picked prediction pipeline with a content-based recommendation engine that's combining weighted credit matching and cosine similarity. These results reflect the available dataset and test split. The anticipated value is an estimate rather than a live-market quote or professional valuation, at least in practice.
Abstract
Heart Disease Prediction Using Machine Learning with Explainable AI
Karthi Balaji M, Nithila K, Sandhya Dillybabu, Dr. Beenarani Manoj
Abstract: Heart disease is a major health issue all over the world and therefore it is important for potential risk patterns to be identified early on. Machine Learning (ML) methods are able to examine clinical data and detect patterns that can assist in putting people into one of two categories: those with heart disease and those without heart disease. The present paper describes a heart disease prediction system based on machine learning that makes use of the Cleveland portion of the UCI Heart Disease dataset. The system applies Logistic Regression and Random Forest classifiers to study patient characteristics such as age, sex, chest pain type, resting blood pressure, cholesterol level, maximum heart rate, and other clinical parameters. Before it is split into training and testing sets, the dataset goes through a preprocessing stage. The models are assessed using accuracy, precision, recall, the F1-score, and the Receiver Operating Characteristic Area Under the Curve (ROC-AUC). In order to enhance interpretability, SHAPley Additive explanations (SHAP) are employed to show the contribution of the input features to the model's predictions. A user interface based on Streamlit allows users to input patient information and see the output from the model. The system integrates prediction, model comparison, and explainability within a single application. The actual performance figures are derived from experimentation and have to be stated after the implementation has been carried out. The system is meant for educational and research use only and should not be used as a replacement for professional medical diagnosis.
Keywords: Heart disease prediction, machine learning, Logistic Regression, Random Forest, SHAP, explainable AI, UCI dataset.
Keywords: Heart disease prediction, machine learning, Logistic Regression, Random Forest, SHAP, explainable AI, UCI dataset.
Abstract
A Survey on SLM for stylized Marathi Prose and Poetry Generation
Manasi Bhole, Riddhi Lahare, Veda Kimbahune, Vedika Kute, P. P. Joshi
Abstract: Text can be generated using Small Language Models with less computational power but it is doubtful that how well literature text can be created using these on low resource Indic Languages. Marathi Language is a good case study for this. While generation good Marathi text using SLMs , not only correct Marathi is required but also the quality of text , cultural fit and uniqueness of author should be depicted in the generated text.
This survey is about understanding how well SLMs can create text , poems , stories given an author style. This survey focuses on recent studies on this. Three main aspects of this survey are as mentioned below 1. SLMs perform good on specific tasks but it doesn’t apply when it comes to low resource language. 2. The quality of literature depends on multiple aspects and automatic evaluation does not assure accuracy. 3. Author style does not indicate copying only
Keywords: they commonly use and SLMs try to capture only surface level feature of author instead of the depth.
This survey is about understanding how well SLMs can create text , poems , stories given an author style. This survey focuses on recent studies on this. Three main aspects of this survey are as mentioned below 1. SLMs perform good on specific tasks but it doesn’t apply when it comes to low resource language. 2. The quality of literature depends on multiple aspects and automatic evaluation does not assure accuracy. 3. Author style does not indicate copying only
Keywords: they commonly use and SLMs try to capture only surface level feature of author instead of the depth.
Abstract
GEOVALUE PREDICTOR HOUSE PRICE PREDICTION FOR INDIA
Dr. Beenarani Manoj, Thanish L.S., Sushaanth.M, Durai Raj.R, Tharun V S
Abstract: Predicting the price of a house can be a complicated task. There are many factors involved in determining the price of property such as location, size, state of property, nearby facilities and type of property. However, there is a lot of variety even between different cities and within cities between different localities in India. This makes it hard to determine whether a property is a fair price based on the information provided.
This paper discusses GeoValue Predictor, a machine learning web application that can predict residential property prices for a selection of cities in India. This takes into account features such as city, locality type, property type, furnishing, number of bedrooms and bathrooms, built up area, floor, age of property, distance to metro stations and city centre, parking, lift and gated community facilities. Four different regression models are compared (Linear Regression, Ridge Regression, Random Forest and Gradient Boosting).
The prototype has been trained on a synthetic data set with 6000 rows. Gradient Boosting gave the highest R squared value (0.960) out of the four models. The mean absolute error was 10.80 lakh rupees and the root mean squared error was 16.10 lakh rupees. We also get an indication of how important each feature is, an indication of how prices vary between different cities, an indication of the predicted price range and an indication of how much it would cost to buy or rent.
This project does not aim to replace professionals who value properties but aims to provide a quick and easy way to give an indication of how much a property may be worth.
This paper discusses GeoValue Predictor, a machine learning web application that can predict residential property prices for a selection of cities in India. This takes into account features such as city, locality type, property type, furnishing, number of bedrooms and bathrooms, built up area, floor, age of property, distance to metro stations and city centre, parking, lift and gated community facilities. Four different regression models are compared (Linear Regression, Ridge Regression, Random Forest and Gradient Boosting).
The prototype has been trained on a synthetic data set with 6000 rows. Gradient Boosting gave the highest R squared value (0.960) out of the four models. The mean absolute error was 10.80 lakh rupees and the root mean squared error was 16.10 lakh rupees. We also get an indication of how important each feature is, an indication of how prices vary between different cities, an indication of the predicted price range and an indication of how much it would cost to buy or rent.
This project does not aim to replace professionals who value properties but aims to provide a quick and easy way to give an indication of how much a property may be worth.
Abstract
Tamil Named Entity Recognition Using Classical Machine Learning, Quantum Feature Processing, and a Simplified QNLP-Inspired Component
Fexi Maxwell G, Paavai Anand G, Sridevi S
Abstract: Tamil Named Entity Recognition (NER) is an important natural language processing task for identifying persons, organizations, and locations in Tamil text. This paper presents a preliminary integrated workflow combining classical machine learning (ML), quantum machine learning (QML)-inspired feature processing, and a simplified quantum natural language processing (QNLP)-inspired component. Tamil text was processed using XLM-RoBERTa, and the resulting representations were reduced using Principal Component Analysis. Classical features, simulated quantum features, and a simplified QNLP-inspired representation were evaluated with logistic-regression classifiers. In the prototype experiment, the classical baseline achieved the highest accuracy and weighted F1-score. The results demonstrate the feasibility of constructing a hybrid workflow, but they do not establish quantum advantage. The quantum circuit used fixed random parameters, and the QNLP component was a simplified proof-of-concept representation.
Keywords: Tamil NLP, Named Entity Recognition, Machine Learning, Quantum Machine Learning, Quantum Natural Language Processing, XLM-RoBERTa
Keywords: Tamil NLP, Named Entity Recognition, Machine Learning, Quantum Machine Learning, Quantum Natural Language Processing, XLM-RoBERTa
Abstract
Intelligent Anomaly Detection for Digital Transactions Using Machine Learning
Desale Harshvardhan, Dr Pradip Patil, Prof Kiran Chaudhary, Namita Devare, Keshar Patil, Divya Jagtap
Abstract: The exponential proliferation of digital payment mechanisms, such as the Unified Payments Interface (UPI), mobile financial wallets, and high-velocity digital payment gateways, has revolutionized consumer transactions while simultaneously creating fertile attack surfaces for automated cyber fraud. Traditional static rule-based security systems suffer from rigidity against zero-day evasion techniques, high false positive rates, and significant operational overhead. In this research, we propose FinGuard AI, an intelligent anomaly detection framework that couples domain-driven mathematical balance invariance with a Two-Phase Hybrid Machine Learning Engine comprising supervised classification, unsupervised spatial anomaly detection, and explainable decision support. Evaluated on a 20,000- transaction synthetic financial benchmark modeled after PaySim and Mendeley financial schemas (19,200 legitimate and 800 confirmed fraudulent transactions), the framework introduces four specialized balance discrepancy features based on double-entry accounting conservation principles. We perform a comparative evaluation across five machine learning archetypes: Logistic Regression (linear baseline), Decision Tree, Random Forest (100 estimators), a 3-layer Deep Neural Network (MLP 128-64-32), and Isolation Forest (unsupervised). The hybrid engine synthesizes posterior class probabilities with normalized Isolation Forest anomaly scores, weighted at 70 percent and 30 percent respectively, subject to domain safety invariant overrides. On the stratified test split of 4,000 transactions, Random Forest and Decision Tree achieved 100.0 percent accuracy (precision, recall, F1 score, and ROC-AUC of 1.000), Logistic Regression achieved 99.92 percent accuracy, Deep Neural Network achieved 99.85 percent accuracy (0.963 recall), and unsupervised Isolation Forest captured 109 out of 160 fraudulent transactions (97.28 percent accuracy, 0.937 ROC-AUC) without ground-truth label exposure. We explicitly analyze why synthetic data generation and balance discrepancy invariants yield near-perfect test separability, document current deployment limitations, and outline critical directions for real-world validation.
Keywords: Digital Transaction Fraud, Anomaly Detection, Machine Learning, Random Forest, Isolation Forest, Hybrid Detection, Financial Security, Explainable AI.
Keywords: Digital Transaction Fraud, Anomaly Detection, Machine Learning, Random Forest, Isolation Forest, Hybrid Detection, Financial Security, Explainable AI.
Abstract
Design and Performance Evaluation of a Post-Quantum Password Manager Using Argon2id, AES-256-GCM, and ML-KEM
Sanjana Pawar, Rama Pandey, Mrs. Soly Zacharish
Abstract: Quantum computing. It also requires that we think about post‑quantum methods to keep sensitive credentials safe for a long time. This paper presents a post‑quantum password manager that uses Argon2id to derive keys, from passwords AES‑256‑GCM to encrypt the vault with authentication and ML‑KEM to set up post‑quantum keys. The versions ML‑KEM‑512, ML‑KEM‑768 and ML‑KEM‑1024 were. Tested under identical conditions using twenty benchmarks. We examined generation, encapsulation, decapsulation, the size of cryptographic objects and how the password vault works. The computational cost and storage space grow as the ML‑KEM parameter set becomes larger. ML‑KEM‑768 turned out to balance security, performance and storage so it was chosen for the password‑manager prototype.
Keywords: Post‑Quantum Cryptography, Password Manager, Argon2id, AES‑256‑GCM, Password Security, Cryptographic Performance
Keywords: Post‑Quantum Cryptography, Password Manager, Argon2id, AES‑256‑GCM, Password Security, Cryptographic Performance
Abstract
CYBERSECURITY AWARENESS AMONG COLLEGE STUDENTS
Ayush Jaiswal, Rishi Jha
Abstract: Cybersecurity awareness is increasingly important for college students who use smartphones, social media, online learning platforms, cloud services and other internet-based applications. This study examines awareness of everyday cybersecurity practices, including password security, phishing recognition, multi-factor authentication, website verification, software updates and privacy settings. A descriptive questionnaire-based research design is presented. The study uses a model sample of 50 respondents and analyses the responses through frequencies, percentages and charts. The findings highlight areas of awareness as well as areas where additional cybersecurity education could be useful. The report concludes with practical recommendations concerning phishing awareness, authentication, passwords, privacy and cybersecurity training.
Keywords: Cybersecurity, Cybersecurity Awareness, College Students, Phishing, Password Security, Multi-Factor Authentication, Online Privacy
Keywords: Cybersecurity, Cybersecurity Awareness, College Students, Phishing, Password Security, Multi-Factor Authentication, Online Privacy
Abstract
Green IT: Reducing the Energy Use and Carbon Footprint of Institutional Computing Infrastructure – A Case Study of a Mid-Size Campus Data Centre
Sahil Patil, Sahil Mohite, Dr. Nisha Yadav
Abstract: Rapid digitalization has made computing one of the fastest-growing sources of electricity demand in institutions, yet most server rooms and computer labs run far below their efficient operating point. This study applies operations research (OR) techniques – energy and utilization analysis, bin-packing optimization, a linear power model and scenario simulation – to (i) analyze the energy and carbon footprint of a representative mid-size campus IT facility, (ii) test Green IT interventions, and (iii) propose an optimal combination of measures. Using a synthetic but realistic dataset of 120 servers, 400 desktop computers and supporting network and cooling equipment, we show that server consolidation, cooling optimization and PC power management can reduce annual electricity use by about 62% and, with a 25% solar offset, cut carbon emissions by about 71%.
Abstract
Detection of Glaucoma using Machine Learning Approach
Zaki Hasan, Prof. Syed Akhtar Hussain, Zafar Ul Hasan
Abstract: Conventional clinical diagnosis is often expensive, time-consuming, dependent on expert interpretation, and not readily available in primary healthcare settings. This study proposes an automated computer-aided diagnosis system for detecting and analysing retinal defects using fundus images. The proposed hybrid framework focuses on multiple anatomical and pathological structures, including retinal blood vessels, optic disc, and exudate lesions. Image pre- processing techniques are employed to enhance retinal images and extract relevant regions of interest. Retinal vessels are segmented using a Fully Connected Conditional Random Field (FC-CRF), followed by supervised classification using SVM and KNN. Optic disc and exudate lesions are segmented using mathematical morphology, fuzzy-set theory, and adaptive thresholding. The proposed system aims to support efficient, automated, and reliable screening of glaucoma and diabetic retinopathy (DR), particularly where limited training data and clinical resources are available.
Keywords: Glaucoma, Fuzzy Logic, Machine Learning, Eyes, Diabetes.
Keywords: Glaucoma, Fuzzy Logic, Machine Learning, Eyes, Diabetes.
Abstract
Intelligent Fraud and Dispute Management Systems
A Sunitha, B Lavanya, B Pallavi, Anita Patil
Abstract: Fraud and dispute complaints in banking are often received through multiple channels such as online banking, mobile applications, customer service, and branch offices. Handling these complaints manually can lead to delays, incorrect case routing, duplicate complaints, inconsistent investigation, and difficulty in tracking case progress. The proposed Intelligent Fraud and Dispute Management System provides a centralized platform to manage fraud and dispute cases in an organized and efficient manner. The system collects customer and transaction details, validates the information, classifies the complaint based on its type, prioritizes cases according to their urgency, identifies possible duplicate cases, and assigns them to the appropriate investigation team. Investigators can access the case details, record investigation findings, update the case status, and take appropriate action based on the available information. The system maintains a complete record of each case from complaint registration to final resolution, allowing bank staff to monitor the progress and customers to receive timely information about their complaints. The proposed system helps streamline the overall case management process, minimize processing delays and errors, improve coordination between teams, and provide a more transparent and reliable way of handling fraud and dispute complaints.
Keywords: Financial Fraud Management, Dispute Management, Machine Learning, TF-IDF, Logistic Regression, Text Classification, Automated Complaint Categorization, Priority-Based Complaint Routing, Investigation Management.
Keywords: Financial Fraud Management, Dispute Management, Machine Learning, TF-IDF, Logistic Regression, Text Classification, Automated Complaint Categorization, Priority-Based Complaint Routing, Investigation Management.
Abstract
Agentic AI-Driven Intrusion Detection and Autonomous Incident Response
Lipsa Dash
Abstract: Limitations associated with the traditional IDS systems which are based on pre-defined rules and human intervention have been highlighted due to the quick rise of the advanced cyber-attacks. Goal-driven software agents which can constantly watch the events in the network, recognize malicious activities, plan for the defence plans and take intelligent reaction actions independent of human intervention are called as agentic artificial intelligence (AI). This novel Agentic AI-Driven Intrusion Detection and Autonomous Incident Response Framework proposed in this paper is comprised of the explainable AI, machine learning, reinforcement learning, threat intelligence and autonomous decision making. In real-time, the proposed framework can detect anomalies, learn adaptive policies, automate threat containment and recover incidents having minimal false positives and reaction time. According to the experimental results, the proposed approach outperforms traditional IDS systems in terms of detection accuracy, precision, response time and resilience. A proactive and intelligent cybersecurity architecture is provided by the framework for the enterprise networks, cloud environments, Internet of Things ecosystem and vital infrastructure that needs continuous protection against new cyber threats.
Keywords: Agentic AI, Intrusion Detection System, Autonomous Incident Response, Cybersecurity, Threat Intelligence, Reinforcement Learning, Explainable AI, Network Security.
Keywords: Agentic AI, Intrusion Detection System, Autonomous Incident Response, Cybersecurity, Threat Intelligence, Reinforcement Learning, Explainable AI, Network Security.
Abstract
A Review of Deep Learning for Real-Time Sign Language Recognition
Minal V. Patil, Pravin I. Patil, Dr. D. M. Patel
Abstract: Communication is the fundamental exchange of thoughts and messages through speech, signals, visuals, and behavior. For deaf and hard-of-hearing individuals, sign language serves as the primary medium for interaction. However, because most non-disabled individuals do not understand sign language, a significant communication barrier exists. This paper presents a user-friendly Human-Computer Interface (HCI) software system designed to translate American Sign Language (ASL) fingerspelling hand gestures into text and speech in real time. The proposed system captures gesture images via a webcam, extracts hand landmarks using the MediaPipe library, and projects them onto a clean white canvas to overcome real-world lighting and background variations. A Convolutional Neural Network (CNN) model is trained on these skeletal representations to recognize individual alphabets (A–Z) and form complete words. The recognized text is then converted into spoken audio using the pyttsx3 library. The proposed approach achieves 97% recognition accuracy under standard conditions and over 99% accuracy in controlled environments.
Keywords: American Sign Language, Convolutional Neural Network, MediaPipe, Human-Computer Interface, Gesture Recognition, Text-to-Speech Conversion.
Keywords: American Sign Language, Convolutional Neural Network, MediaPipe, Human-Computer Interface, Gesture Recognition, Text-to-Speech Conversion.
Abstract
A Novel Hybrid Dynamic Pheromone Optimization Algorithm with Entropy-Guided Learning and Historical Route Intelligence for the Travelling Salesman Problem
Mrs. M. Hemalatha, Dr. N. Kamaraj
Abstract: The Travelling Salesman Problem (TSP) remains one of the most extensively studied NP-hard combinatorial optimization problems due to its wide range of applications in logistics, transportation, manufacturing, and network design. Although Ant Colony Optimization (ACO) has demonstrated promising performance for solving TSP, traditional pheromone updating mechanisms often suffer from premature convergence, loss of population diversity, and stagnation in local optima. This paper proposes Hybrid Dynamic Pheromone Optimization (HDPO), a novel optimization algorithm that integrates entropy-guided learning, adaptive pheromone regulation, historical route intelligence, and dynamic local search mechanisms into a unified framework. HDPO dynamically adjusts pheromone evaporation and reinforcement based on search diversity, route entropy, and solution improvement rates. Furthermore, an elite historical memory structure preserves high-quality route knowledge across generations, while a stagnation recovery mechanism reintroduces exploration when convergence becomes excessive. Experimental evaluations on benchmark instances from TSPLIB can demonstrate that HDPO achieves superior solution quality, faster convergence, and enhanced robustness compared with classical ACO variants. The proposed method offers an effective balance between exploration and exploitation, making it suitable for large-scale TSP instances and other combinatorial optimization problems.
Keywords: Community Network, Travelling Salesman Problem (TSP), Ant Colony Optimization (ACO), HDPO, TSPLIB.
Keywords: Community Network, Travelling Salesman Problem (TSP), Ant Colony Optimization (ACO), HDPO, TSPLIB.
Abstract
Neocube Realty: A Web-based approach for comprehensive real estate Business System
Pranjal Rane, Muskan Madhwani, Vaishnavi Sawarkar, Kajal Gawande, Chanchal Karwa, Achal Pande, Prof. S.A. Solanke
Abstract: The real estate sector faces challenges in fragmented property management, manual inquiry processing, and inefficient lead allocation. This paper presents Neocube Realty, a centralized web-based platform developed using React, Spring Boot, Spring Security, and MySQL. The system implements RESTful APIs, role-based access control (RBAC), ORM, and workflow automation for integrated property and transaction management. An automated inquiry- to-lead conversion mechanism and round-robin lead assignment algorithm distribute leads efficiently among active brokers. The platform further incorporates advanced property filtering, geospatial mapping, Point-of-Interest (POI) analysis, and video-based property visualization, including customer-provided property videos collected during the listing process, to enhance property discovery and decision making. The system supports the complete lead lifecycle, including broker assignment, status tracking, site visits, bookings, and deal management, through role-specific interfaces and analytical dashboards. The proposed system demonstrates the integration of workflow automation, geospatial intelligence, and immersive visualization into a unified and scalable real estate management platform.
Keywords: Real Estate Management, React, Spring Boot, REST API, RBAC, MySQL, Round-Robin Algorithm, Geospatial Intelligence, POI Analysis, Blueprint Visualization.
Keywords: Real Estate Management, React, Spring Boot, REST API, RBAC, MySQL, Round-Robin Algorithm, Geospatial Intelligence, POI Analysis, Blueprint Visualization.
Abstract
Fake News Detection Using Machine Learning and Natural Language Processing
Dhanashri Satish Deshmukh, Ms Rucha Mohite
Abstract: The use of websites social media and online news platforms has made it possible to receive information quickly at the same time the amount of misleading and false news available online has increased checking every news article manually can be time-consuming this study therefore explores a machine learning-based method for automatically classifying news articles the proposed system uses the t fake news separate collections of fake and real news articles the news text is cleaned and converted into using term frequency-inverse document frequency tf-idf these features are given to a logistic regression classifier for binary assigns each article fake or real category on the test data the developed model obtained an accuracy of 9882 the results indicate that the selected combination of text performs well on in study keywords fake news detection machine learning natural language processing tf-idf logistic regression
Abstract
FARMER E-MARKET FOR BETTER PRODUCT DEMAND PREDICTION
Revati Vishwas Shitole, Dr. Shveti Chandan
Abstract: Farmers often have to decide what to grow and how much to bring to the market without having a clear idea of future demand. When too much of a crop reaches the market, some of it may remain unsold and prices may fall. When the available quantity is too low, buyers may struggle to find the product they need. This paper proposes a Farmer E-Market system that brings farmers and market information together and uses past demand records to estimate future demand for selected crops. The proposed system allows a farmer to enter the crop name, quantity and expected harvest date. It then compares the farmer’s planned supply with current supply, expected future supply and predicted demand for the selected region and period. The first version uses a Random Forest regression model trained on historical crop-demand records containing year, month, region, crop and market demand. The project uses a time-based train-test split so that earlier records are used for training and later records are used for testing. The system is intended to provide a simple, understandable view of demand and supply rather than guarantee a sale or a particular price. Its usefulness will depend on the quality, coverage and freshness of the data available.
Keywords: Farmer E-Market, Crop Demand Prediction, Random Forest, Supply-Demand Comparison, Digital Agriculture.
Keywords: Farmer E-Market, Crop Demand Prediction, Random Forest, Supply-Demand Comparison, Digital Agriculture.
Abstract
AI-BASED RANSOMWARE DETECTION FRAMEWORK USING MACHINE LEARNING
Kajal Shildar Padvi, Dr.Rucha Mohite
Abstract: Ransomware remains one of the major cybersecurity threats because modern ransomware variants continuously evolve and may differ considerably from previously identified samples. Traditional signature-based detection approaches are effective against known threats but may have limited capability against previously unseen variants. Machine learning provides an alternative approach by learning discriminative patterns from malicious and benign executable samples. However, high classification performance under a conventional random train-test split does not necessarily demonstrate generalization to ransomware families that were not represented during training.
This research proposes an AI-Based Ransomware Detection Framework Using Machine Learning based on static characteristics of Windows Portable Executable (PE) files. The framework extracts a compact set of eight static features consisting of API count, DLL count, unique API count, unique DLL count, file size, packed status, entropy, and PE timestamp year. Five machine learning algorithms—Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM—are trained and compared using accuracy, precision, recall, F1score, ROC-AUC, false positive rate, and false negative rate.
The experimental framework additionally incorporates validation-based probability threshold analysis, ransomware- family-disjoint evaluation, family-level generalization analysis, Isolation Forest-based novelty detection, and SHAP- based explainability. Random Forest achieved the best overall standard-test performance with 95.42% accuracy, 96.22% precision, 94.28% recall, 95.24% F1-score, and 98.98% ROC-AUC. When evaluated on ransomware families excluded from training, the Random Forest model achieved 90.20% accuracy, 98.88% precision, 87.02% recall, 92.57% F1-score, and 97.72% ROC-AUC. The family-disjoint experiment demonstrates that detection becomes more challenging when the model encounters ransomware families not represented during training. Isolation Forest additionally identified approximately 8.04% of unseen ransomware samples as anomalous relative to the known ransomware distribution.
The results demonstrate that a compact static PE representation combined with ensemble machine learning can provide strong ransomware classification performance while enabling additional analysis of generalization, novelty, and interpretability.
Keywords: Ransomware Detection, Machine Learning, Static Analysis, Portable Executable, Random Forest, Explainable AI, SHAP, Unseen Ransomware Families, Anomaly Detection, Malware Detection.
This research proposes an AI-Based Ransomware Detection Framework Using Machine Learning based on static characteristics of Windows Portable Executable (PE) files. The framework extracts a compact set of eight static features consisting of API count, DLL count, unique API count, unique DLL count, file size, packed status, entropy, and PE timestamp year. Five machine learning algorithms—Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM—are trained and compared using accuracy, precision, recall, F1score, ROC-AUC, false positive rate, and false negative rate.
The experimental framework additionally incorporates validation-based probability threshold analysis, ransomware- family-disjoint evaluation, family-level generalization analysis, Isolation Forest-based novelty detection, and SHAP- based explainability. Random Forest achieved the best overall standard-test performance with 95.42% accuracy, 96.22% precision, 94.28% recall, 95.24% F1-score, and 98.98% ROC-AUC. When evaluated on ransomware families excluded from training, the Random Forest model achieved 90.20% accuracy, 98.88% precision, 87.02% recall, 92.57% F1-score, and 97.72% ROC-AUC. The family-disjoint experiment demonstrates that detection becomes more challenging when the model encounters ransomware families not represented during training. Isolation Forest additionally identified approximately 8.04% of unseen ransomware samples as anomalous relative to the known ransomware distribution.
The results demonstrate that a compact static PE representation combined with ensemble machine learning can provide strong ransomware classification performance while enabling additional analysis of generalization, novelty, and interpretability.
Keywords: Ransomware Detection, Machine Learning, Static Analysis, Portable Executable, Random Forest, Explainable AI, SHAP, Unseen Ransomware Families, Anomaly Detection, Malware Detection.
Abstract
Bridging the Language Gap: NLP Based Translation of Government Schemes for Marathi- Speaking Farmers
Pooja Suryabhan Gaygaye, Ms.Yogita Kadbane
Abstract: Government provide many schemes for farmer.the scheme is on official portal,but most of the scheme are in English .In Maharashtra many Farmers know only Marathi language,they did not aware about English language,due to this they are not able to take the benefits of the governments schemes.
My Research objectives is ,all farmers can get benefited from governments schemes,they can get advantage of the government policies also. I am used Experimental Methodology and I am used A Transformer-based Neural Machine Translation .the performance of model was evaluated using BLEU and CHRF metrics.the system helps to farmers for easily understand government scheme information,specially for Marathi speaking farmers.In future I am focusing on expanding the Data set of farmer related schemes,supporting additional Indian languages and voice based interaction.
Keyword: Natural Language Processing(NLP),English-Marathi Translation ,Transformer,Neural Machine Translation,Government Schemes,Agricultural Information,Marathi Language
My Research objectives is ,all farmers can get benefited from governments schemes,they can get advantage of the government policies also. I am used Experimental Methodology and I am used A Transformer-based Neural Machine Translation .the performance of model was evaluated using BLEU and CHRF metrics.the system helps to farmers for easily understand government scheme information,specially for Marathi speaking farmers.In future I am focusing on expanding the Data set of farmer related schemes,supporting additional Indian languages and voice based interaction.
Keyword: Natural Language Processing(NLP),English-Marathi Translation ,Transformer,Neural Machine Translation,Government Schemes,Agricultural Information,Marathi Language
Abstract
SIX-YEAR ENROLLMENT TREND ANALYTICS AT WESLEYAN UNIVERSITY-PHILIPPINES: A DATA-DRIVEN BASIS FOR REGISTRAR'S OFFICE DIGITALIZATION AND AUTOMATION
Yolanda M. Javier, Shirley DC. Ramos, Abigail C. Bautista, Carmeliza S. Valdez, Rozen Jake DR. Punay, Gemmuel M. Borja, John Andrei J. Ignacio
DOI: 10.17148/IJARCCE.2026.151001
Abstract: This study analyzed institutional enrollment trends across six academic years, from Academic Year (AY) 2018–2019 to AY 2023–2024, with emphasis on changes by college and sex. Specifically, it compared the combined enrollment counts during the first three academic years (2018–2019 to 2020–2021) with those during the last three academic years (2021–2022 to 2023–2024). A descriptive-comparative approach was used to analyze institutional enrollment records. Frequency counts, percentages, numerical differences, and percentage changes were employed to identify patterns of enrollment growth and decline. Findings revealed that annual enrollment initially decreased by 5.34% from AY 2018–2019 to AY 2019–2020 but subsequently recovered, with the largest annual increase of 26.97% occurring in AY 2021–2022. Enrollment reached a six-year high of 15,043 in AY 2023–2024. Comparing the two three-year periods, overall enrollment increased from 29,901 to 42,231, representing an increase of 12,330 or 41.24%. Seven of the eight colleges recorded enrollment growth, while the College of Education (COED) decreased by 3.73%. The College of Nursing (CON) demonstrated the largest percentage increase at 157.23%, followed by the College of Arts and Sciences (CAS) at 44.15% and the College of Allied Medical Sciences (CAMS) at 41.05%. Female enrollment remained higher than male enrollment, accounting for 61.30% of enrollment during the last three-year period. Female enrollment increased by 48.81%, compared with 30.70% among males. However, sex composition differed across colleges, with males remaining the majority in CCJE and CECT. The findings demonstrate substantial institutional enrollment expansion, although the magnitude and sex distribution of growth varied across academic units. These enrollment analytics provide an empirical and data-driven basis for institutional planning and for advancing the digitalization and automation of registrar operations, particularly in enrollment monitoring, automated reporting, data visualization, and decision support. Integrating historical enrollment data into a digital registrar information environment may facilitate more efficient monitoring of enrollment patterns, resource allocation, staffing, student services, and program development. The figures represent enrollment counts across the first semester, second semester, and summer terms and should not be interpreted as counts of unique individual students.
Keywords: academic programs, enrollment trends, higher education, institutional planning, sex distribution
Keywords: academic programs, enrollment trends, higher education, institutional planning, sex distribution
Abstract
Enhanced Sentiment-Based Stock Market Recommendation System Using Machine Learning
Sanapala Ramesh
DOI: 10.17148/IJARCCE.2026.151002
Abstract: Stock prices are influenced not only by historical price movement but also by public sentiment expressed through financial news. Retail investors usually consult charting platforms and news platforms separately, with no automated bridge between them. This project presents an implemented, Flask-based web application that combines quantitative technical analysis of stock prices with machine-learning based sentiment analysis of financial news to produce a transparent BUY, HOLD or SELL recommendation.
Historical price data is sourced from Yahoo Finance and used to calculate the 20-day moving average, the 14-period Relative Strength Index (RSI) and the Moving Average Convergence Divergence (MACD). Financial news is gathered from NewsAPI, cleaned with NLTK, and converted into TF-IDF and hand-crafted sentiment features. Three classifiers, namely Random Forest, Linear Support Vector Machine and Multinomial Naive Bayes, are trained, and the best is selected by test-set accuracy. Article-level predictions are aggregated into a daily Sentiment Index, which is combined with a price-trend score using the weighted formula 0.6 x Sentiment + 0.4 x Trend. An RSI-based safety filter prevents extreme-market signals from being followed blindly. All data and results are stored in SQLite and displayed on an interactive Chart.js dashboard.
The system was run on 411 stored news articles covering seven US and Indian stocks. On an 80/20 split, the Linear SVM achieved the highest test accuracy (100.00%), followed by Random Forest (98.80%) and Naive Bayes (79.52%), and was selected automatically. Because the sentiment labels are generated by a financial keyword lexicon rather than by human annotators, these metrics measure agreement with that labelling heuristic and not validated real-world sentiment accuracy. This limitation is noted explicitly throughout the report.
Keywords: stock market, sentiment analysis, technical indicators, RSI, MACD, TF-IDF, Random Forest, SVM, Naive Bayes, Flask.
Historical price data is sourced from Yahoo Finance and used to calculate the 20-day moving average, the 14-period Relative Strength Index (RSI) and the Moving Average Convergence Divergence (MACD). Financial news is gathered from NewsAPI, cleaned with NLTK, and converted into TF-IDF and hand-crafted sentiment features. Three classifiers, namely Random Forest, Linear Support Vector Machine and Multinomial Naive Bayes, are trained, and the best is selected by test-set accuracy. Article-level predictions are aggregated into a daily Sentiment Index, which is combined with a price-trend score using the weighted formula 0.6 x Sentiment + 0.4 x Trend. An RSI-based safety filter prevents extreme-market signals from being followed blindly. All data and results are stored in SQLite and displayed on an interactive Chart.js dashboard.
The system was run on 411 stored news articles covering seven US and Indian stocks. On an 80/20 split, the Linear SVM achieved the highest test accuracy (100.00%), followed by Random Forest (98.80%) and Naive Bayes (79.52%), and was selected automatically. Because the sentiment labels are generated by a financial keyword lexicon rather than by human annotators, these metrics measure agreement with that labelling heuristic and not validated real-world sentiment accuracy. This limitation is noted explicitly throughout the report.
Keywords: stock market, sentiment analysis, technical indicators, RSI, MACD, TF-IDF, Random Forest, SVM, Naive Bayes, Flask.
Abstract
Automated Multi-Student Attendance Using Video-Based Facial Recognition and Embedding Matching
Likhitha Alajangi, Dr. Kuda Nageswara Rao
DOI: 10.17148/IJARCCE.2026.151003
Abstract: A major aspect of determining student attendance is having accurate records; unfortunately, traditional attendance systems are inefficient, not accurate, and can be manipulated for ”proxy” attendance. A way around these obstacles is to create an intelligent, automated attendance method using deep learning facial recognition technology based on a pretrained model. In addition to traditional still-image approaches, our system begins with recording a short video of each participating student, leading to frame-by-frame extraction for constructing a robust and diverse dataset containing facial image samples of each student. This new pipeline from video to image will improve the representation of students’ faces by using the three ways students’ faces are captured (pose, expression, and lighting) to provide a more realistic photo of each student. In this approach, we will utilise a pretrained deep learning face recognition algorithm to extract the distinctive features of student faces from our generated dataset. A single classroom image containing multiple students will be used for attendance. The system will identify faces in the group image and extract their unique features by using three face detection methods, namely Haar Cascade, MTCNN, and InsightFace, which are evaluated and compared. For recognition, FaceNet and ArcFace embedding models are investigated using cosine similarity matching to perform both extraction and recognition, which helps us generate face embeddings for each person, compare those features to the database using cosine similarity, and mark attendance for recognized students with no manual intervention. This approach should successfully connect individual-level training data with group-level testing conditions, allowing for accurate facial recognition in the classroom.
Keywords: Automated Attendance System, Facial Recognition, Pretrained Deep Learning Model, Video Frame Extraction.
Keywords: Automated Attendance System, Facial Recognition, Pretrained Deep Learning Model, Video Frame Extraction.
Abstract
AI-Based Semantic Understanding of Code Mixed Tamil-English Text in Informal Digital Communication
Aditya Raman, Sushmita Kundu, Aditya Balaji, G.Paavai Anand
DOI: 10.17148/IJARCCE.2026.151005
Abstract: Tamil-English code-mixing, popularly known as “Tanglish,” is a pervasive mode of informal digital communication among Tamil speakers, frequently rendered in Romanized script across social media and messaging platforms. Existing natural language processing (NLP) systems, largely trained on monolingual or formally structured corpora, are not optimized to interpret the semantic, sentiment, and intent-level nuances of such hybrid, non-standardized text. This paper proposes a comparative research framework to evaluate the capability of different computational approaches—traditional NLP techniques, BERT-based transformer models, and large language models (LLMs)—in understanding Tamil-English code-mixed text, alongside a proposed preprocessing pipeline designed to normalize Romanized Tamil variants prior to model inference. We outline a methodology for constructing an annotated code-mixed dataset, detail the architecture of each comparative model, and define evaluation metrics spanning accuracy, F1-score, and semantic similarity. The study aims to establish whether targeted preprocessing can meaningfully improve semantic fidelity for low-resource, code-mixed language understanding, contributing toward AI systems that better preserve and process minority and regional language expression in digital spaces.
Keywords: code-mixing, Tanglish, Tamil-English NLP, Romanized Tamil, semantic preservation, low-resource language processing, transformer models
Keywords: code-mixing, Tanglish, Tamil-English NLP, Romanized Tamil, semantic preservation, low-resource language processing, transformer models
Abstract
Comparative Analysis of Convolutional Neural Network Architectures for Automated Waste Classification
Sakshi Anil Parkhe
DOI: 10.17148/IJARCCE.2026.151006
Abstract: Effective waste classification is a critical component of modern smart waste management systems, enabling automated sorting and recycling at scale. This paper presents a comprehensive comparative study of six convolutional neural network (CNN) architectures (VGG16, ResNet50, MobileNetV3, EfficientNet-B0, AlexNet, and a custom- designed CNN) for the task of automated waste classification into six categories: cardboard, glass, metal, paper, plastic, and trash. Each model is evaluated on a standardized dataset of 2,527 images under identical experimental conditions, including data augmentation, learning rate scheduling, and transfer learning where applicable. Performance is assessed using accuracy, precision, recall, F1-score, receiver operating characteristic (ROC) curves, and precision-recall (PR) curves. Experimental results demonstrate that the custom CNN achieves the highest classification accuracy of 95.05% with a macro-average ROC-AUC of 0.9974, outperforming established architectures. ResNet50 and MobileNetV3 follow closely at 94.07% and 93.87%, respectively, offering favorable trade-offs between accuracy and computational complexity. The analysis further includes per-class performance evaluation, confusion matrix analysis, and computational efficiency comparison, providing actionable insights for deploying waste classification systems in resource-constrained environments.
Keywords: waste classification, convolutional neural networks, transfer learning, deep learning, image classification, recycling, smart waste management
Keywords: waste classification, convolutional neural networks, transfer learning, deep learning, image classification, recycling, smart waste management
Abstract
MACHINE LEARNING FRAMEWORK FOR DETECTION OF MENTAL STRESS
Archana Borse, MS. Yogita Kadbane
DOI: 10.17148/IJARCCE.2026.151007
Abstract: Mental stress is becoming a common issue among students and individuals due to factors such as workload, lack of sleep, screen time, deadline pressure, financial concerns, personal responsibilities, and limited relaxation time. Identifying stress levels can be difficult because different factors can affect each individual differently. This research presents a machine learning framework for the detection of mental stress using questionnaire-based data. Data was collected from 51 respondents through a Google Form and included factors such as sleep hours, screen time, workload, physical activity, socialization, sleep quality, deadline pressure, financial concerns, personal or family responsibilities, and social support. The collected data was processed and used to calculate a stress score, based on which respondents were classified into Low, Moderate, and High stress levels. The dataset contained 17 columns and the stress-level distribution included 10 Low, 37 Moderate, and 4 High stress responses. A machine learning classification model was developed using Python to identify stress-level patterns. The model achieved an accuracy of approximately 81.82%. The study also used data visualization to understand the distribution of stress and the relationship between different factors and stress levels. The proposed framework shows how data analysis and machine learning can be used as a preliminary approach for detecting and understanding mental stress patterns.
Keywords: Mental Stress, Machine Learning, Stress Detection, Stress Level, Data Analysis, Python, Classification
Keywords: Mental Stress, Machine Learning, Stress Detection, Stress Level, Data Analysis, Python, Classification
Abstract
HemaSight AI: A Four-Stage Deep Learning Pipeline for Automated Leukemia Detection and Post-Transplant Monitoring
Varun Doddagoudar, Vishwa, Shashank, Madhushree M
DOI: 10.17148/IJARCCE.2026.151008
Abstract: Leukemia is a life-threatening hematological malignancy where early and accurate detection significantly improves patient survival. Conventional diagnosis relies on manual microscopic examination of peripheral blood smears, a process that is time-consuming, observer-dependent, and prone to error. This paper presents HemaSight AI, a four- stage artificial intelligence pipeline for automated leukemia screening and post-transplant monitoring. Stage 1 applies XGBoost to classify complete blood count (CBC) parameters, including 17+ extended markers such as SGOT, SGPT, creatinine, potassium, chimerism, and GVHD status. Stage 2 employs a conditional generative adversarial network (cGAN) to synthesize microscopic blood smear images alongside a ResNet18-based similarity engine that computes a weighted four-metric match (semantic, color, structural, pixel). Stage 3 uses linear regression on longitudinal blood test data to detect early signs of relapse before clinical symptoms appear. Stage 4 integrates Google Gemini via a multilingual chatbot supporting seven languages. The platform is implemented using Next.js, FastAPI, Supabase, and PyTorch, deployed entirely on free-tier infrastructure. Evaluation across 68 test cases achieved a 98.5% pass rate with sub-3-second API response times. Results demonstrate that a low-cost, scalable, and clinically accessible AI tool can be built without proprietary infrastructure.
Keywords: Leukemia detection, XGBoost, conditional GAN, ResNet18, time-series analysis, large language model, medical AI, multilingual chatbot.
Keywords: Leukemia detection, XGBoost, conditional GAN, ResNet18, time-series analysis, large language model, medical AI, multilingual chatbot.
Abstract
Smart Irrigation System for Precision Farming
Mr. Bhagyesh N Chaudhari, Dr. Dinesh D Puri
DOI: 10.17148/IJARCCE.2026.151010
Abstract: Smart Irrigation System for Precision Farming is an IoT-based and sensor-driven approach designed to improve water management, crop productivity, and agricultural sustainability. Conventional irrigation practices often depend on fixed schedules or manual decisions, which can result in over-irrigation, under-irrigation, water wastage, and increased labour requirements. The proposed system integrates soil-moisture, temperature, humidity, and rainfall sensors with a microcontroller such as Arduino or ESP32 to continuously monitor field conditions in real time. Based on the collected data, the system automatically determines the irrigation requirement and controls water pumps or solenoid valves accordingly. Weather information can also be incorporated to avoid unnecessary irrigation during expected rainfall and to adapt watering according to environmental conditions. IoT connectivity enables farmers to remotely monitor soil and weather parameters through a web or mobile dashboard. The system provides a low-cost, scalable, and farmer-friendly solution that supports precision agriculture by delivering the right amount of water at the right time. Such intelligent irrigation can conserve water and energy, reduce manual effort, improve crop health, and contribute to sustainable agricultural development.
Abstract
Multi-Crop Disease Detection and Management Recommendation System Using Deep Learning
K. Hemalatha, Prof. K. Venkata Rao, P. Swathi
DOI: 10.17148/IJARCCE.2026.151011
Abstract: Crop diseases are a major challenge in agriculture, as late or incorrect identification can cause crop damage and reduce productivity. Existing crop disease detection systems mainly focus on identifying diseases from plant leaf images, but provide limited support for disease management. This creates a gap between disease identification and practical management. The proposed Multi-Crop Disease Detection and Management Recommendation System Using Deep Learning addresses this gap by combining disease detection with management recommendations. The system uses deep learning and transfer learning models to classify crop diseases from leaf images. It provides chemical and organic management approaches based on the identified crop and disease. Thus, the system connects disease identification with suitable management practices, making it more useful for practical crop disease management.
Keywords: Crop Disease Detection, Deep Learning, Management Recommendation System, Multi-Crop Classification, Transfer Learning.
Keywords: Crop Disease Detection, Deep Learning, Management Recommendation System, Multi-Crop Classification, Transfer Learning.
Abstract
Digital Document Verification System
Dr. S.S. Kokila, Sriniva.S
DOI: 10.17148/IJARCCE.2026.151012
Abstract: Digital documents are widely used in education, banking, healthcare, government, employment, and business. Ensuring the authenticity and integrity of these documents has become increasingly important due to the growing number of forged, altered, and fraudulent documents. Traditional verification methods rely heavily on manual inspection and comparison with original records, which can be time-consuming, costly, and prone to human errors. Recent developments in Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and Optical Character Recognition (OCR) have enabled automated document verification with improved accuracy and efficiency. Machine learning algorithms such as Random Forest, Support Vector Machine, Decision Trees, Logistic Regression, and K-Nearest Neighbors can classify genuine and fraudulent documents using extracted features. Deep learning models such as Convolutional Neural Networks (CNNs) can analyze document images and detect manipulation and forgery. This review examines recent advancements in digital document verification, including OCR, image processing, digital signatures, blockchain, biometric verification, cloud computing, and multimodal document analysis. The study also discusses commonly used datasets, verification methods, evaluation metrics, challenges, research gaps, and future research directions. The findings indicate that integrating AI-based verification, secure digital signatures, blockchain, and multimodal analysis can support the development of reliable, secure, scalable, and efficient digital document verification systems
Keywords: Digital Document Verification, Document Authentication, Machine Learning, Deep Learning, Optical Character Recognition, Artificial Intelligence, Blockchain, Digital Signature, Forgery Detection.
Keywords: Digital Document Verification, Document Authentication, Machine Learning, Deep Learning, Optical Character Recognition, Artificial Intelligence, Blockchain, Digital Signature, Forgery Detection.
Abstract
CUSTOMER COMPLAINT TRACKING SYSTEM
Dr. L. Baby Victoria, Abinaya D
DOI: 10.17148/IJARCCE.2026.151013
Abstract: Customer complaints provide valuable information about problems in products, services, delivery, support, and customer experience. However, organizations often receive a large number of complaints through websites, email, social media, call centers, and mobile applications, making manual processing slow and difficult to manage. A customer complaint tracking system can improve this process by automatically collecting, classifying, prioritizing, routing, monitoring, and resolving complaints. This paper presents a review of recent approaches based on Natural Language Processing (NLP), Machine Learning (ML), Deep Learning (DL), sentiment analysis, topic modelling, and transformer-based models for intelligent complaint management. NLP techniques can convert unstructured complaint text into useful information, while machine learning algorithms can classify complaints according to category, urgency, department, or severity. Sentiment analysis can identify customer emotions and support priority assignment, whereas automated routing can forward complaints to the appropriate department. Recent research has also explored BERT and AI-enabled complaint-handling systems. The review identifies important research gaps related to multilingual complaints, limited labelled datasets, explainability, privacy, real-time tracking, integration with existing business systems, and human-AI collaboration. Future systems should combine NLP, explainable AI, large language models, sentiment analysis, automated escalation, and real-time dashboards to provide faster and more transparent complaint resolution.
Keywords: Customer Complaint Tracking, Complaint Management, Natural Language Processing, Machine Learning, Sentiment Analysis, Text Classification, BERT, Topic Modelling, Automated Routing, Customer Service.
Keywords: Customer Complaint Tracking, Complaint Management, Natural Language Processing, Machine Learning, Sentiment Analysis, Text Classification, BERT, Topic Modelling, Automated Routing, Customer Service.
Abstract
Crop Recommendation System for Farmers Using Convolutional Neural Network
D. Savitha, Gayathri U, Priya S
DOI: 10.17148/IJARCCE.2026.151014
Abstract: Agriculture plays an important role in food security and the economic development of many countries. Selecting the most suitable crop for a particular field is a major challenge for farmers because crop growth depends on soil properties, nutrients, temperature, humidity, rainfall, pH, and climatic conditions. Traditional crop selection mainly depends on farmers experience and local knowledge, which may not always provide accurate recommendations under changing environmental conditions. Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) provide effective solutions for data-driven agricultural decision making. This paper presents a review of crop recommendation systems with particular emphasis on Convolutional Neural Networks (CNNs). CNNs can automatically learn useful patterns from soil images, remote-sensing images, and other agricultural data, while CNN- based hybrid models can also be combined with environmental and soil parameters. Recent studies have explored Random Forest, XGBoost, Support Vector Machine, LSTM, CNN, attention-based networks, Transformer models, and hybrid deep-learning architectures for crop recommendation. This review examines research published from 2020 to 2026, summarizes their datasets, preprocessing methods, algorithms, and major findings, and identifies research gaps related to dataset availability, regional generalization, explainability, multimodal data integration, and real-time deployment. A CNN-based crop recommendation system can potentially provide farmers with localized and intelligent recommendations while reducing inappropriate crop selection, improving productivity, and supporting sustainable farming.
Keywords: Crop Recommendation, Convolutional Neural Network, CNN, Deep Learning, Precision Agriculture, Smart Farming, Soil Classification, Machine Learning, Farmers, Agriculture.
Keywords: Crop Recommendation, Convolutional Neural Network, CNN, Deep Learning, Precision Agriculture, Smart Farming, Soil Classification, Machine Learning, Farmers, Agriculture.
Abstract
Equilibrium Dynamics and Threshold Conditions in Tumor–Immune Models with Chemotherapy
Shivangi Chauhan*, Prof. Diwari Lal
DOI: 10.17148/IJARCCE.2026.151015
Abstract: This research presents a non-linear tumor-immune-chemotherapy model that distinguishes drug sensitive from drug resistant tumors with respect to chemotherapy dynamics and immune response. The model includes Gompertzian tumor growth, immune mediated killing, differential drug sensitivity, treatment induced drug resistance, immune suppression, and drug clearance. Biologically feasible equilibrium states were established by analytical investigation along with biological conditions of threshold for invasion by sensitive or resistant tumor cell types. Chemotherapy is demonstrated to have a double-edged sword effect in this model; increasing treatment intensity will be effective at suppressing sensitive tumor cells, however it also promotes the relative proportion of resistant tumor cells. Numerically simulated results demonstrate that local stability remains around positive equilibria and resistant tumor cells can remain present within the model despite an increase in chemotherapy. Further sensitivity analysis was conducted to identify factors that influence the models behavior including tumor growth rate, immune killing rates, chemotherapy efficacy, selective pressure on resistance development, degree of immune suppression and drug clearance. Ultimately the data supports that reduction in sensitive tumor population through chemotherapy does not eliminate all cancer (resistant cells), nor prevent recurrence due to continued presence of resistant cells.
Keywords: Tumor–immune dynamics, chemotherapy, drug resistance, mathematical modeling, equilibrium analysis, threshold conditions, Gompertz growth, resistance invasion
Keywords: Tumor–immune dynamics, chemotherapy, drug resistance, mathematical modeling, equilibrium analysis, threshold conditions, Gompertz growth, resistance invasion
Abstract
Quantum Assisted Predictive Maintenance and Scheduling System for Industry 4.0
Muppuri Satish Kumar
DOI: 10.17148/IJARCCE.2026.151016
Abstract: Predictive Maintenance is one of the primary applications of Industry 4.0 for detecting potential equipment failures in advance of unplanned outages. With the increasing number of Industrial Internet of Thing(IoT) sensors available today, there is now the ability to access multi-modal data from machines that can be used to monitor conditions and predict failures. Traditional methods have been limited in their ability to capture the complex relationship between the different sensor measurements and make decisions regarding maintenance based upon limited resources. A quantum- assisted framework for integrating Machine Failure Prediction and Maintenance Scheduling. A Hybrid Quantum Neural Network (HQNN) will be utilized to analyze the different types of industrial sensor data and determine the likelihood of failure. The quantum component of our model will use a parameterized quantum circuit to represent the analyzed sensor data as quantum states; then measure these quantum states to produce a classical representation of the probability of failure. Once we have determined this probability of failure, it will be combined with the priority of the machine and its maintenance needs to create a maintenance scheduling problem. A QAOA-based method will be developed to develop a multi-shift maintenance schedule while considering all available resources.
Keywords: Industry 4.0, Predictive Maintenance, Hybrid Quantum Neural Network (HQNN), Parameterized Quantum Circuit (PQC), CNOT Ring Entanglement, QAOA, QUBO, Shift Scheduling
Keywords: Industry 4.0, Predictive Maintenance, Hybrid Quantum Neural Network (HQNN), Parameterized Quantum Circuit (PQC), CNOT Ring Entanglement, QAOA, QUBO, Shift Scheduling
Abstract
Chronic Kidney Disease Prediction System Using Machine Learning
Savitha. D, Haritha. V P, Mathisha. N
DOI: 10.17148/IJARCCE.2026.151017
Abstract: Chronic kidney disease (CKD) is a progressive condition that often remains unnoticed until a large part of kidney function has been lost, which makes early screening one of the most effective ways of delaying dialysis and transplantation. Machine learning (ML) can learn the relationship between routine clinical and laboratory measurements and the presence of CKD, and can turn it into a fast, low-cost decision-support tool. This paper presents a structured
and ensemble classifiers, feature selection, class-imbalance handling, explainable artificial intelligence and progression prediction. The studies are compared in a summary table and the common gaps are identified: reliance on small public datasets, near-perfect but weakly validated accuracy, limited external validation, explanations that never reach the user, and few complete systems that connect prediction to patient records and reports. Based on these gaps, the motivation, application areas and future directions for a practical web-based CKD prediction system built on a Random Forest classifier are discussed.
Keywords: Chronic kidney disease, Machine learning, Randomforest, Early prediction, Explainable AI, Feature selection, Clinical decision support.
and ensemble classifiers, feature selection, class-imbalance handling, explainable artificial intelligence and progression prediction. The studies are compared in a summary table and the common gaps are identified: reliance on small public datasets, near-perfect but weakly validated accuracy, limited external validation, explanations that never reach the user, and few complete systems that connect prediction to patient records and reports. Based on these gaps, the motivation, application areas and future directions for a practical web-based CKD prediction system built on a Random Forest classifier are discussed.
Keywords: Chronic kidney disease, Machine learning, Randomforest, Early prediction, Explainable AI, Feature selection, Clinical decision support.
Abstract
Women Enterprise Hub: A Web-Based Platform for Women Entrepreneurs
D. Savitha, S. Nitharsana, S. Subarna
DOI: 10.17148/IJARCCE.2026.151018
Abstract: Women Enterprise Hub is a web-based platform designed to support women entrepreneurs, particularly those involved in home-based and small-scale sectors such as handicrafts, clothing, jewellery and food products. Many women entrepreneurs depend on informal networks, local markets and scattered social media channels to promote their businesses, while information about government schemes, funding opportunities, mentoring and training is distributed across different sources. The proposed system brings these activities together through a centralized portal. The platform provides user registration and authentication, business profile management, product listing, marketplace functions, order processing, delivery estimation, government scheme information, mentor discovery and session booking, training information, reviews and suggestions. A database-driven architecture is used to manage users, businesses, products, orders and supporting services. The system aims to provide a structured digital environment that improves visibility, access to support information and interaction between women entrepreneurs, mentors and customers.
Keywords: Women Entrepreneurship, Web Portal, Marketplace, Mentoring, Government Schemes, PHP, MySQL
Keywords: Women Entrepreneurship, Web Portal, Marketplace, Mentoring, Government Schemes, PHP, MySQL
Abstract
DESIGN AND DEVELOPMENT OF AN
ESP-BASED SMART MOBILE, MANIPULATOR FOR REMOTE OBJECT, HANDLING, Kuncham Preetham, Dr. K. Raja Kumar
DOI: 10.17148/IJARCCE.2026.151020
Abstract: The integration of robotic mobility and object manipulation provides a practical solution for remote robotic operations. This work presents a manually controlled mobile robotic platform integrated with a servo-based robotic arm for basic object handling. An ESP32 microcontroller coordinates the DC motors, servo motors, and ultrasonic sensing according to user commands and sensed distance. The vehicle supports forward, backward, left, right, and stop operations, while the robotic arm provides base rotation, shoulder and elbow movement, and gripper open/close functions. A wireless control interface provides remote operation, and an ultrasonic sensor is used for obstacle detection. Experimental demonstration shows successful vehicle movement, obstacle stopping, arm actuation, and pick- and-place operation. The resulting platform is compact, portable, economical, and suitable for educational, remote- handling, and basic mobile robotic applications.
Keywords: ESP32, mobile manipulator, robotic arm, ultrasonic sensor, wireless control, servo motor, obstacle detection, pick-and-place.
Keywords: ESP32, mobile manipulator, robotic arm, ultrasonic sensor, wireless control, servo motor, obstacle detection, pick-and-place.
Abstract
An Intelligent Risk-Adaptive Framework for Secure Examination Paper Distribution, Traceability, and Early Leak Detection
Mr. Swapnil S. Chaudhari, Asst. Prof. Sapana A. Fegade
DOI: 10.17148/IJARCCE.2026.151021
Abstract: Examination-paper leakage threatens the fairness, credibility, and operational continuity of academic and competitive examinations. Conventional protection mechanisms rely heavily on controlled physical handling, restricted access, and manual supervision; digital distribution reduces some transportation risks but introduces new requirements for encryption, authentication, key management, traceability, and monitoring. This research presents a risk-adaptive examination-security framework that integrates four complementary capabilities: secure digital distribution, tamper- evident chain-of-custody tracking, semantic leak detection, and risk-based security control selection. The study follows a design-science and applied-experimental approach. The proposed architecture separates a Java/Spring Boot security core from a Python-based natural-language-processing service. Question papers are protected through encryption, controlled release, copy-level fingerprinting, and auditable access events. Suspected public-source leaks are represented as a semantic-search problem in which original questions are compared with noisy, partial, paraphrased, OCR-derived, and multilingual variants using sentence embeddings. A risk engine maps examination characteristics to an interpretable risk score and recommends stronger controls for higher-risk examinations. The empirical evidence retained from the study is a pilot analysis of 66 documented Indian examination-paper-leak incidents from 2014-2024. Logistic Regression and Random Forest were evaluated using five-fold stratified cross-validation against a majority-class baseline. The baseline achieved an F1-score of 0.842, compared with 0.674 for Logistic Regression and 0.739 for Random Forest, indicating that the small incident dataset and limited features are insufficient for reliable prediction of incident- confirmation status. This classification task is treated as a pilot study rather than as evidence that future examination risk can already be predicted reliably. The manuscript therefore defines a reproducible experimental protocol for evaluating semantic leak detection and the complete risk-adaptive system without fabricating unperformed results. A working prototype implementing all four modules - the Java/Spring Boot security core with role-based access control, the SHA- 256 hash-linked chain of custody, a Python microservice offering both a TF-IDF baseline and Sentence-BERT similarity, and the risk-scoring engine - was built and verified through fourteen automated checks on the core security logic and an end-to-end run confirming encryption, tamper-evident logging, and tamper detection. The principal contribution is an integrated and testable research architecture, evidenced by a working prototype, that explicitly connects prevention, traceability, detection, and adaptive response, together with an identification of the data and validation requirements needed before such a system can be deployed in a real examination environment.
Keywords: Examination Security; Question-Paper Leakage; Risk-Adaptive Security; Semantic Similarity; Sentence- BERT; Blockchain; Chain of Custody; Digital Watermarking; NLP; Secure Distribution.
Keywords: Examination Security; Question-Paper Leakage; Risk-Adaptive Security; Semantic Similarity; Sentence- BERT; Blockchain; Chain of Custody; Digital Watermarking; NLP; Secure Distribution.
Abstract
Smart Classroom Seat Allocation System Using IoT, Artificial Intelligence, Machine Learning and Optimization Techniques: A Review
Dr. P. Anitha, V Karishni
DOI: 10.17148/IJARCCE.2026.151022
Abstract: The transformation of conventional classrooms into intelligent learning environments has created new opportunities for automating classroom management and improving student learning experiences. Among these challenges, student seat allocation is an important task because seating arrangements can influence accessibility, classroom interaction, student engagement, learning conditions and effective utilization of available space. Traditional seat allocation is predominantly performed manually and becomes increasingly complex when the number of students, classrooms and allocation constraints increases. A Smart Classroom Seat Allocation System can address these limitations by integrating the Internet of Things (IoT), Artificial Intelligence (AI), Machine Learning (ML) and optimization techniques. This paper presents a comprehensive review of research published between 2020 and 2026 concerning classroom seating, intelligent seat allocation, smart classroom technologies, IoT-based monitoring and optimization-based allocation approaches. Existing studies have explored genetic algorithms, mathematical programming, graph-based heuristics, agent-based modelling, sociometric sensing, RFID, IoT devices and automated classroom management systems. Research findings indicate that seating position and proximity can affect student engagement, academic performance, peer interaction and social relationships. IoT technologies can further provide real-time information regarding seat occupancy and classroom conditions, enabling dynamic allocation and monitoring. However, existing approaches frequently address individual aspects of the problem, such as examination seating, social distancing, classroom interaction or occupancy monitoring, rather than providing an integrated allocation framework. Therefore, there is a significant opportunity to develop an adaptive system that combines student characteristics, classroom constraints, seat availability, accessibility requirements, learning preferences and real-time occupancy information within a unified optimization framework. This review examines existing approaches, identifies their limitations and research gaps and discusses potential applications and future directions for developing intelligent, flexible and scalable smart classroom seat allocation systems.
Keywords: Smart Classroom, Seat Allocation, Artificial Intelligence, Machine Learning, Internet of Things, Optimization, RFID, Student Management, Smart Education.
Keywords: Smart Classroom, Seat Allocation, Artificial Intelligence, Machine Learning, Internet of Things, Optimization, RFID, Student Management, Smart Education.
Abstract
PERSONALISED HEALTHCARE RECOMMENDATION SYSTEM USING XGBOOST AND SHAP
Savitha. D, Sowmiya. V, Akshaya. C
DOI: 10.17148/IJARCCE.2026.151023
Abstract: The advancement of Artificial Intelligence (AI) and Machine Learning (ML) has brought significant improvements to healthcare by supporting disease prediction, individual health assessment, and clinical decision- making. However, conventional healthcare systems often provide generalized recommendations without adequately considering individual differences in demographic information, lifestyle, physiological conditions, medical history, and health risks. This review proposes an AI-Based Personalized Healthcare Recommendation System that integrates Extreme Gradient Boosting (XGBoost) with SHapley Additive exPlanations (SHAP). The approach combines the strong predictive performance of XGBoost with Explainable Artificial Intelligence (XAI) to generate personalized and understandable healthcare insights. The system analyzes various healthcare attributes, including demographic details, body measurements, vital signs, lifestyle factors, medical history, family history, laboratory results, and existing health conditions to assess potential disease risks. XGBoost is effective for structured healthcare data as it can capture complex nonlinear relationships between different health-related features. SHAP provides explanations by identifying the contribution of individual features to each prediction, enabling users and healthcare professionals to understand important factors such as blood glucose, body mass index, cholesterol levels, physical activity, sleep duration, and family history.
Keywords: Personalized Healthcare, Machine Learning, XGBoost, SHAP
Keywords: Personalized Healthcare, Machine Learning, XGBoost, SHAP
Abstract
Real-Time Sign Language Translator Using Video Prediction
Geetha. N, Mohanashankare.V.K, Prathanyaa.K
DOI: 10.17148/IJARCCE.2026.151025
Abstract: Sign language is a visual language in which meaning is conveyed through hand shape, movement, body posture, facial expression, and temporal relationships between gestures. Automatic translation from video to text can reduce communication barriers between sign-language users and people who do not know the language. This paper presents a real-time sign language translator using video prediction, in which a webcam or uploaded video is processed as a sequence rather than as independent images. The proposed framework combines video acquisition, frame preprocessing, hand and pose landmark extraction, spatial feature learning, temporal sequence modelling, sign prediction, text generation, and optional speech synthesis. OpenCV can be used for frame handling, while MediaPipe provides efficient landmark extraction. CNN(Convolutional Neural Network), GRU(Gated Recurrent Unit), LSTM( Long Short-Term Memory), or Transformer models can then learn spatial and temporal patterns in the video. Recent research from 2020 to 2026 shows a transition from static and isolated-sign recognition toward continuous sign- language recognition and end-to-end sign-to-text translation. However, practical deployment remains challenging because of signer variation, occlusion, variable signing speed, limited continuous datasets, non-manual cues, temporal alignment, and computational latency. The proposed research direction therefore emphasizes efficient video prediction, robust temporal modelling, and real-time deployment, with particular relevance to Indian Sign Language.
Keywords: Sign Language Recognition, Video Prediction, Computer Vision, Deep Learning, MediaPipe, Temporal Modeling, Transformer, Indian Sign Language, Real-Time Translation, Speech Synthesis.
Keywords: Sign Language Recognition, Video Prediction, Computer Vision, Deep Learning, MediaPipe, Temporal Modeling, Transformer, Indian Sign Language, Real-Time Translation, Speech Synthesis.
Abstract
Agent-less Windows System Vulnerability and Network Scanner for Security Assessment and Reporting
Satvik Dilip Barhate, Dr. Dinesh Puri
DOI: 10.17148/IJARCCE.2026.151026
Abstract: Modern Windows endpoints and small enterprise networks expose multiple security-relevant services, software versions, configurations, and network paths. Conventional vulnerability assessment approaches may require endpoint agents, extensive deployment, or separate tools for network discovery and host-level inspection. This research presents CyberWolf, an agent-less Windows system vulnerability and network scanner designed to combine network discovery, service enumeration, structured scan-result processing, and rule-based security assessment in a lightweight workflow. The proposed approach uses Nmap for network discovery and XML output as a machine-readable data source, followed by parsing, normalization, security checks, finding prioritization, and report generation. The design is intended for controlled and authorized assessment of Windows hosts without installing a permanent software agent on each target. The research defines a repeatable assessment lifecycle, a normalized data model, example security rules, evaluation dimensions, limitations, and future extensions. The work is positioned as a defensive assessment framework for identifying network-exposed conditions and supporting remediation decisions.
Keywords: Windows Security, Vulnerability Assessment, Agent-less Scanning, Network Discovery, Nmap, CyberWolf, Security Automation, Vulnerability Management
Keywords: Windows Security, Vulnerability Assessment, Agent-less Scanning, Network Discovery, Nmap, CyberWolf, Security Automation, Vulnerability Management
Abstract
Healthcare Insurance Fraud Detection Using Machine Learning and Deep Learning
Siddanathi Prasanth Srinivas Sai
DOI: 10.17148/IJARCCE.2026.151027
Abstract: Healthcare insurance fraud causes significant financial losses and is difficult to detect due to the large volume and complexity of healthcare claims. This paper presents an integrated fraud detection framework using beneficiary, inpatient, outpatient, and provider-level fraud-label data. The datasets are cleaned, integrated, and transformed into provider-level features, with 558,211 inpatient and outpatient claim records forming the unified claims dataset. XGBoost, Random Forest, and TabNet are employed to identify providers associated with potential fraudulent activity, while class imbalance is addressed during model development. A local Llama 3.2-based explanation layer generates human-readable explanations, and a SHA-256 hash-chained audit ledger provides tamper-evident records of detection activities. A dashboard is also used to present prediction results and model performance. The proposed framework integrates fraud detection, explainability, and secure auditing for effective healthcare insurance fraud analysis.
Keywords: Healthcare Insurance Fraud Detection, TabNet, XGBoost, Random Forest, Ensemble Learning, Explainable Artificial Intelligence (XAI), Llama 3.2, SHA-256 Hash-Chained Audit Ledger.
Keywords: Healthcare Insurance Fraud Detection, TabNet, XGBoost, Random Forest, Ensemble Learning, Explainable Artificial Intelligence (XAI), Llama 3.2, SHA-256 Hash-Chained Audit Ledger.
Abstract
Grouping-Aware Numerical Verification for Offline Retrieval-Augmented Question Answering over Institutional Documents
Sudheer Katuri, G. Sharmila Sujatha
DOI: 10.17148/IJARCCE.2026.151028
Abstract: Institutions hold large volumes of policy documents, reports and spreadsheets that are still searched manually. Retrieval-augmented generation (RAG) offers a natural interface to such collections, but mainstream deployments route confidential material through hosted language-model APIs and provide no guarantee that figures quoted in an answer actually occur in the source text. This paper presents an offline RAG system for institutional document question answering in which ingestion, embedding, retrieval and generation all execute on commodity hardware with no external network dependency, and in which every numeral in a generated answer is verified against the retrieved context before the answer is shown. The central observation is that naive numeric extraction — the digit-matching regular expression used in comparable pipelines — fragments digit-grouped quantities such as 45,23,000 into unrelated components, causing a verifier built on it to reject faithful answers. We describe a grouping-aware extraction and verification procedure that handles Indian and Western grouping conventions, decimal and percentage forms, and enumeration markers. On a 15case extraction benchmark the proposed extractor is exact on 15/15 cases against 5/15 for the baseline, and substituting the baseline extractor into the verifier introduces false rejections of faithful answers. Measured retrieval latency over an exhaustive 384dimensional index remains below one millisecond to roughly 5,000 chunks and grows linearly to about 10.5 ms at 50,000 chunks, which bounds the corpus size at which exact search remains appropriate.
Keywords: retrieval-augmented generation, hallucination detection, offline inference, vector search, document question answering, privacy-preserving systems.
Keywords: retrieval-augmented generation, hallucination detection, offline inference, vector search, document question answering, privacy-preserving systems.
Abstract
Genuine Co-Movement, Spurious Trend, and a Ratio That Does Not Help: Variable Construction for Nigerian Macroeconomic Models
Ofierohor Ufuoma Earnest
DOI: 10.17148/IJARCCE.2026.151029
Abstract: Macroeconomic models routinely include exchange-rate and trade variables side by side as separate regressors. This paper examines whether they are separate, using 67 quarterly Nigerian observations from 2008Q2 to 2024Q4, and reports two results that point in different directions. First, the co-movement among the exchange rate, exports, imports and total trade is genuine rather than a shared trend. Levels correlations run from 0.95 to 0.99, and they survive first differencing at 0.68 to 0.93, so this is not a spurious-regression artefact. Second, the co-movement between these series and GDP is spurious: the level correlation between the exchange rate and GDP is 0.607 and falls to -0.132 in differences, and trade against GDP falls from 0.599 to -0.106. This matters for a common variable-construction choice. Normalising trade by GDP is often justified as removing shared scale, but here the numerator and denominator have no genuine co- movement, and the available series compounds the problem by dividing nominal trade by real GDP. The resulting ratio reaches 189% of GDP and tests as I(2), which invalidates the ARDL bounds procedure. We recommend testing correlations in differences before treating a ratio as a solution to collinearity, and checking that numerator and denominator share a price basis.
Keywords: Multicollinearity; Spurious Correlation; Trade Openness; Variable Construction; Nigeria.
Keywords: Multicollinearity; Spurious Correlation; Trade Openness; Variable Construction; Nigeria.
Abstract
Abhimanyu Chakravyuha: An Evidence-Aware AI Framework for Fraud Intelligence Through Semantic Analysis, Mutation Detection, and Graph-Based Reasoning
MD Auranzeb Khan, Kunal, Srijan Mani Tripathi, Vyom Pandey, Dr.G Paavai Anand
DOI: 10.17148/IJARCCE.2026.151030
Abstract: Fraudulent activities increasingly involve evolving narratives, altered textual claims, interconnected entities, and coordinated patterns that cannot always be identified by examining individual records in isolation. This paper presents Abhimanyu Chakravyuha, a proposed AI-powered fraud intelligence framework integrating natural language processing, semantic similarity analysis, mutation detection, graph-based relationship analysis, and evidence-aware verification.
The framework processes textual claims and related records, retrieves semantically relevant evidence, identifies potential modifications to previously observed patterns, and represents relationships using a graph structure. An evidence- verification component evaluates whether available information supports, contradicts, or fails to establish a claim. An abstention mechanism allows the system to return an inconclusive result when evidence is insufficient or conflicting. The design emphasizes evidence traceability, uncertainty management, and human oversight. It separates semantic similarity from factual support, suspicious patterns from confirmed fraud, and automated analysis from human-reviewed decisions. This paper specifies the architecture, methodology, and experimental design; quantitative performance is not claimed without experimental validation.
Keywords: Fraud Intelligence, Natural Language Processing, Semantic Similarity, Claim Verification, Mutation Detection, Knowledge Graphs, Explainable AI, Evidence Retrieval, Abstention, Uncertainty Calibration
The framework processes textual claims and related records, retrieves semantically relevant evidence, identifies potential modifications to previously observed patterns, and represents relationships using a graph structure. An evidence- verification component evaluates whether available information supports, contradicts, or fails to establish a claim. An abstention mechanism allows the system to return an inconclusive result when evidence is insufficient or conflicting. The design emphasizes evidence traceability, uncertainty management, and human oversight. It separates semantic similarity from factual support, suspicious patterns from confirmed fraud, and automated analysis from human-reviewed decisions. This paper specifies the architecture, methodology, and experimental design; quantitative performance is not claimed without experimental validation.
Keywords: Fraud Intelligence, Natural Language Processing, Semantic Similarity, Claim Verification, Mutation Detection, Knowledge Graphs, Explainable AI, Evidence Retrieval, Abstention, Uncertainty Calibration
Abstract
AI-DRIVEN CLINICAL DECISION SUPPORT SYSTEM FOR INTELLIGENT MEDICAL REPORT ANALYSIS AND DISEASE PREDICTION
Sravanthi Duvvuri, Dr. G. Sharmila Sujatha
DOI: 10.17148/IJARCCE.2026.151031
Abstract: The AI-Driven Clinical Decision Support System for Intelligent Medical Report Analysis and Disease Prediction enhances healthcare decision-making by integrating Artificial Intelligence into clinical data processing. The system uses Natural Language Processing (NLP) to extract symptoms from patient complaints, suggests relevant medical tests, and applies machine learning algorithms to analyze laboratory reports and detect abnormal values. It predicts possible heart, liver, and kidney diseases with probability scores, identifies the key factors behind each prediction, and generates structured clinical summaries to assist doctors in faster evaluation. The heart disease module is built on the merged UCI Heart Disease dataset containing 920 records and 14 attributes. The prototype, developed in Python with spaCy, scikit-learn, SQLite, and Streamlit, demonstrates the complete workflow from patient registration to an explained prediction. Designed as a decision-support tool, the system reduces diagnostic workload, improves patient prioritization, and supports timely intervention while ensuring that final medical decisions remain under professional supervision.
Keywords: Clinical decision support system, electronic health records, natural language processing, machine learning, disease prediction, explainable AI, SHAP, clinical summary.
Keywords: Clinical decision support system, electronic health records, natural language processing, machine learning, disease prediction, explainable AI, SHAP, clinical summary.
Abstract
AI-Enabled Structural Vibration Prediction for Enhanced Stealth Performance
Harish Kumar Kuppili, Kuda Nageswara Rao
DOI: 10.17148/IJARCCE.2026.151033
Abstract: Increasing engineering complexity and the need for dependable structural health monitoring have created a demand for accurate and computationally efficient vibration-prediction methods. Conventional approaches—including physics-based models, extensive experimental testing across operating conditions, and computationally intensive simulations—can be difficult to apply when structures and service environments vary widely and decisions must be made under uncertainty. This study presents an AI-based framework for predicting structural vibration responses using regression algorithms. Five approaches—artificial neural network (ANN), k-nearest neighbours (KNN), decision tree (DT), gradient boosting regression (GBR), and random forest regression (RFR)—were evaluated using historical and simulated data representing material characteristics, stiffened geometries, and applied loading conditions. The comparison indicates that ensemble methods, particularly RFR and GBR, provide stronger predictive accuracy and generalisation than the other evaluated models. GBR delivered the best test performance, while RFR achieved a closely comparable result. ANN could represent nonlinear relationships but required additional computational effort and tuning, whereas KNN and DT were more sensitive to variation in the data. The findings indicate that AI-based regression can support faster preliminary assessment, reduce design-cycle time, and contribute to structural monitoring. This work demonstrates the potential of ensemble learning as a scalable approach to vibration prediction and as a foundation for data-driven infrastructure monitoring, proactive maintenance, anomaly detection, and improved structural safety.
Keywords: Artificial intelligence, vibration prediction, structural health monitoring, predictive models, data-driven methods, machine learning.
Keywords: Artificial intelligence, vibration prediction, structural health monitoring, predictive models, data-driven methods, machine learning.
Abstract
Contract Sentinel: A Literature Survey on Transformer-Based Clause Extraction and Risk Detection in Legal Contracts
Vaishnavi Karpe, Tanishka Makode, Maria Shaikh, P. J. Jambhulkar
DOI: 10.17148/IJARCCE.2026.151034
Abstract: Legal contracts such as non-disclosure agreements, lease agreements, and employment contracts often contain complex clauses that can create substantial financial, legal, and operational risk. Manually reviewing these documents takes time and requires domain expertise that individuals, startups, or small businesses do not always have, leaving them vulnerable to missing unfavourable or unusual terms. In this project, we present Contract Sentinel, an AI system that automatically reviews legal contracts and flags risky or unusual provisions for manual review. The system is based on natural language processing and machine learning algorithms trained on the CUAD (Contract Understanding Atticus Dataset), consisting of 510 real-world contracts with more than 13,000 expert labels across 41 clause types. The system takes a contract document as input, extracts clauses from it, and assigns each clause to a known risk category such as indemnification, liability caps, termination rights, non-compete terms, and auto-renewal conditions. Unusual or unfair clauses are marked with a description of the danger in plain English, allowing users to rapidly grasp and negotiate complex agreements without the need for legal expertise. Contract Sentinel uses transformer-based language models and a curated legal clause taxonomy to speed up and improve contract review, making it more accessible and reliable by reducing the need for expensive legal consultation for routine document checks.
Keywords: Legal NLP, Contract Review, Clause Classification, Risk Detection, CUAD, Legal-BERT, Transformers, Explainable AI
Keywords: Legal NLP, Contract Review, Clause Classification, Risk Detection, CUAD, Legal-BERT, Transformers, Explainable AI
Abstract
SR-Screen-AI: An Explainable Rule-Based AI-Assisted Decision Support System for Human-in-the-Loop Systematic Literature Review Screening
Sruthi R, Rajeev P R
DOI: 10.17148/IJARCCE.2026.151035
Abstract: Systematic literature review screening requires assessment of bibliographic records against predefined criteria. This paper presents SR-Screen-AI, an explainable rule-based AI-assisted decision-support system integrating import, duplicate detection, weighted screening, explanations, reviewer validation, and reporting. Eight software tests passed, and a 50-record baseline trial achieved 81.82% precision, 90.00% recall, and 85.71% F1-score, supporting transparent reviewer-controlled screening.
Keywords: Artificial intelligence, Decision support systems, Explainable artificial intelligence, Human-in-the-loop, Literature screening, Rule-based systems, Systematic literature review
Keywords: Artificial intelligence, Decision support systems, Explainable artificial intelligence, Human-in-the-loop, Literature screening, Rule-based systems, Systematic literature review
Abstract
How Many Normal Images Are Needed? Training-Free Few-Shot Anomaly Detection on Scanning Electron Microscope Images of Integrated Circuits
K. Susmitha Reddy, G. Lavanya Devi
DOI: 10.17148/IJARCCE.2026.151036
Abstract: Anomaly detection on scanning electron microscope (SEM) images of integrated circuits is an essential element in the semiconductor manufacturing quality control process to ensure that no defective chip is delivered. Because SEM data is specific to each product and imaging tool, inspection models are usually built or tuned for each production line. Developing these models involves a large quantity of images and takes considerable effort to collect the data and refine the model until it works reliably. Also, most of the time, collecting data results in a disproportionate amount of normal images compared to anomalous images. Vision foundation models, pretrained on very large image collections, can describe images in great detail and have become strong tools for finding anomalies in photographs. Even so, the best anomaly detection models on the public MIIC dataset have been trained on 23,888 normal images, which is still a large number. In this research, we studied training-free few-shot anomaly detection on MIIC using frozen vision foundation backbones to assess whether their features transfer to SEM images. We extracted features from k normal reference images to form a memory bank, and assigned an anomaly score to each test feature based on its distance to the nearest stored reference feature in the memory bank. We tested six backbones with four values of k and three random seeds. DINOv2- B with 64 references yielded 94.53% image AUROC, which was better than full-data PatchCore (93.33%), while DINOv3 and CLIP were less transferable. A layer analysis revealed that averaging the last four transformer layers played a critical role in localization, raising the AUPRO of DINOv3-S to 95.2% compared with 76.2% for the last layer alone. When the 16 most similar normal references were retrieved for every test image, AUPRO reached 99.23%, nearly equivalent to full-data CFA (99.04%), and the nearest-neighbour search was 129 times quicker than searching through a 2,000-image pool. However, image-level AUROC (94.78%) lagged behind full-data CFA (99.01%). A decision layer calibrated solely on normal images rejected only 0.39% of normal images and detected every anomaly covering at least 0.25% of the image.
Keywords: Few-shot anomaly detection, foundation models, defect inspection, semiconductor manufacturing, electron microscopy.
Keywords: Few-shot anomaly detection, foundation models, defect inspection, semiconductor manufacturing, electron microscopy.
Abstract
A Nightshade-SWIN Transformer-Based Framework for Enhanced Multi-Crop Plant Disease Classification
Dr. Barkha Joshi
DOI: 10.17148/IJARCCE.2026.151037
Abstract: Nightshade crops (Solanaceae) face significant yield threats from phenotypically similar foliar diseases that can be difficult to distinguish during the early stages of infection. Accurate and timely disease identification is therefore essential for reducing crop losses, improving disease management, and supporting sustainable agricultural practices. While Convolutional Neural Networks (CNNs) have demonstrated strong performance in image-based plant disease classification due to their ability to extract spatial and local features, they may have limitations in capturing long-range dependencies and global contextual relationships within leaf images. These limitations become particularly important when disease symptoms exhibit subtle differences in texture, color, shape, and lesion distribution or when images are captured against complex field backgrounds. To address these challenges, we propose Nightshade-SWIN, an optimized deep learning framework based on the Swin Transformer architecture. The proposed framework utilizes a hierarchical shifted-window self-attention mechanism to effectively capture both minute lesion textures and global leaf morphology. By progressively learning multi-scale representations, Nightshade-SWIN can identify discriminative pathological characteristics while maintaining computational efficiency. The model was evaluated on a curated dataset of 5,000 multi- crop images collected under variable field conditions, including differences in illumination, background complexity, leaf orientation, and disease severity. Experimental results demonstrate that Nightshade-SWIN achieves a peak classification accuracy of 94.6%, outperforming ResNet50 and EfficientNet-B0 by 4.9% and 3.4%, respectively. These results indicate the effectiveness of transformer-based feature learning for distinguishing visually similar foliar diseases. Furthermore, interpretability analysis using LIME and Grad-CAM confirms that the model primarily focuses on biologically relevant pathological regions, such as lesions, discoloration, spots, and infected leaf areas. This enhances the transparency and reliability of the proposed framework and supports its potential application in precision agriculture, automated crop monitoring, and early disease diagnosis.
Keywords: Nightshade Crops, Leaf Disease Detection, SWIN-Transformer, Grad-CAM and LIME, Deep Learning
Keywords: Nightshade Crops, Leaf Disease Detection, SWIN-Transformer, Grad-CAM and LIME, Deep Learning
Abstract
A Secure and Privacy-Aware Hybrid Framework for Digital Credential Verification
Jayashri Digambar Sable, Swaranjali Dinesh Shelar, More N.N
DOI: 10.17148/IJARCCE.2026.151040
Abstract: In most organizations, identity is still checked against a central database. This is convenient to run. The catch is that one party ends up owning the records, and people must keep uploading the same documents every time they sign up for something new. In this paper we propose a hybrid framework that uses a blockchain for one job only: keeping a shared record of trust. Each issuer and holder gets a decentralized identifier (DID), and the claims go inside verifiable credentials (VCs). For every credential, a smart contract stores just its hash and its status, so nothing personal is written on the chain. We describe the architecture, follow a credential from issue to revocation, give an eight-step verification algorithm, and discuss the threats and a test plan. We have not built the prototype yet, so no performance numbers appear here; we will report them after we have measured them.
Keywords: blockchain, decentralized identifiers, digital identity, privacy, self-sovereign identity, verifiable credentials
Keywords: blockchain, decentralized identifiers, digital identity, privacy, self-sovereign identity, verifiable credentials
Abstract
An AI-Based Intelligent System for Early Urban Health Risk Assessment
Mali Rajashri Savata, Salunkhe Samruddhi Mahadev, Lokare Sujata Sharad, Ghalame Mayuri Kishor, Linge Sandeep Ganpat
DOI: 10.17148/IJARCCE.2026.151041
Abstract: Rapid urbanization and changing environmental conditions can affect public health in many ways. Factors such as temperature, humidity, rainfall, and air quality may increase the risk of different diseases. This research proposes an AI-Based Community Health Risk Assessment System to predict possible health risks in a community. The proposed system collects real-time and historical environmental data and uses Machine Learning to analyze the data and predict health risks. Based on the prediction, the system can provide early warnings and help authorities take preventive actions. The system also includes health risk hotspot detection feature to identify high risk areas in city based on environmental conditions and predicted health risk. In addition, the system collects combined symptoms reports from citizens and collects them with environmental data and health risk prediction. Using this information, the system generates AI based community health advisories to provide useful health warnings and guidance Overall, the system combines health risk prediction, early warning, hotspot detection, and health advisories in a single platform. It can help health authorities identify high-risk areas, take preventive measures, and make better data-based decisions to improve community health and safety.
Keywords: Artificial Intelligence, Machine Learning, Community Health, Health Risk Prediction, Environmental Data, Hotspot Detection, Early Warning System, Citizen Symptom Reports, Smart City, Public Health.
Keywords: Artificial Intelligence, Machine Learning, Community Health, Health Risk Prediction, Environmental Data, Hotspot Detection, Early Warning System, Citizen Symptom Reports, Smart City, Public Health.
Abstract
Self-Evolving AI Agent Framework for Predicting and Preventing Zero-Day Cyberattacks in a Multi-Cloud Environment Using Federated Learning and Explainable AI
Samruddhi Santosh Gawade, Gayatri Gunwant Gayaki, Ms. Ghodake G.K
DOI: 10.17148/IJARCCE.2026.151042
Abstract: The proposed framework uses intelligent AI agents to continuously monitor cloud activities, detect unusual patterns, and identify potential security threats at an early stage. Federated learning enables different cloud environments to work together in training AI models without sharing their raw security data, thereby helping to maintain data privacy. Explainable AI makes the model’s decisions easier to understand by providing clear reasons behind its predictions, allowing security professionals to investigate suspicious activities and take appropriate action. Additionally, the self- evolving component is designed to adapt to emerging threats by learning from new information and changing attack patterns.
Overall, the framework aims to strengthen threat detection, protect sensitive information, and improve transparency in multi-cloud environments. However, its actual performance needs to be verified through practical implementation and testing with realistic datasets and simulated cyberattacks. Future research should focus on evaluating important factors such as detection accuracy, false-positive rate, response time, and communication overhead to determine the framework’s effectiveness and suitability for real-world cloud security applications.
Keywords: Zero-day cyberattacks, artificial intelligence, self-evolving AI agents, multi-cloud security, federated learning.
Overall, the framework aims to strengthen threat detection, protect sensitive information, and improve transparency in multi-cloud environments. However, its actual performance needs to be verified through practical implementation and testing with realistic datasets and simulated cyberattacks. Future research should focus on evaluating important factors such as detection accuracy, false-positive rate, response time, and communication overhead to determine the framework’s effectiveness and suitability for real-world cloud security applications.
Keywords: Zero-day cyberattacks, artificial intelligence, self-evolving AI agents, multi-cloud security, federated learning.
Abstract
School Management System for Dutuwewa Maha Vidyalaya
Keshika Rathakrishnan, Parahera Nekathige Kaumini Viduthala
DOI: 10.17148/IJARCCE.2026.151043
Abstract: This report presents the design and implementation of a comprehensive School Management System for Dutuwewa Maha Vidyalaya, a streamlined, user-friendly platform aimed at improving efficiency and communication within the school. Existing manual systems often focus on singular functionalities like attendance or grade taking, but there is a growing need for an integrated approach that addresses multiple administrative, academic, and communicative needs. The School Management System provides a centralized platform for school administrators, teachers, students, and parents to seamlessly manage and interact with key aspects of school life, including student enrolment, attendance, event management, resource allocation, and communication. With tools to support notifications, grade selection, and attendance taking, the system promotes increased transparency and engagement between the school and its community. By offering an intuitive interface and robust features, the system enhances operational efficiency, improves communication, and contributes to the school's overall success in meeting academic and administrative goals
Keywords: School Management System, Administrative automation, educational communication, Digital Transformation.
Keywords: School Management System, Administrative automation, educational communication, Digital Transformation.
Abstract
A Voice-Powered Intelligent Healthcare Platform for Elderly Patients with Dementia and Alzheimer's Care
Shinde Akshata Kumar, Pathan Ekhara Mushtak, Parchande Sanika Harishchandra, Khilare Sakshi Kalidas, Sawant Namdev Manik
DOI: 10.17148/IJARCCE.2026.151044
Abstract: Cognitive impairments, including Alzheimer's disease and dementia, represent rapidly expanding global health challenges that steadily erode patient memory, reasoning, and autonomy [1][3]. Standard caregiving frameworks rely on constant human oversight, an approach that is financially demanding and susceptible to delayed emergency interventions. This study introduces a Voice-Driven Smart Healthcare System that merges automated speech processing, voice processing, AI-assisted behavioral analysis, and continuous monitoring to assist elderly users. The architecture allows patients to utilize conversational commands to obtain medication prompts, daily scheduling support, and emergency trigger mechanisms. Simultaneously, an interface supplies caregivers with instantaneous activity records and health metrics. The system's methodology converts clinical necessities into three distinct operational elements: a Voice Processing Engine, an AI-Driven Health Assistance Module, and a Caregiver Administration Dashboard, all supported by a secure cloud-based architecture designed with privacy, consent, and role-based access considerations. Evaluations of feasibility indicate that this solution is practically deployable in domestic and assisted-care environments, offering a more economical alternative to round-the-clock manual supervision. The proposed platform is anticipated to boost medication compliance, accelerate emergency reactions, and alleviate the strain on caregivers, all while safeguarding the privacy of users with minimal digital literacy [9][14]. Future enhancements will target wearable device integration, multi- language conversational capabilities, and predictive risk modeling [15]. Ultimately, by integrating natural voice interfaces with AI analytics and real-time oversight, this initiative seeks to transition dementia management from reactive supervision to a model that promotes dignity and proactive independence.
Keywords: Voice-Driven Healthcare, Dementia Management, Alzheimer's Disease, Voice Interaction , Live Health Tracking
Keywords: Voice-Driven Healthcare, Dementia Management, Alzheimer's Disease, Voice Interaction , Live Health Tracking
Abstract
MetaSociety: A Simulation-Driven AI Policy Advisor for Data-Grounded Smart Governance
Priyanka Jadhav, Payal Dongre, Priti Chaugule, Mayuri Kulkarni
DOI: 10.17148/IJARCCE.2026.151045
Abstract: Policy decisions in local governance are frequently taken with limited empirical grounding, relying on precedent, intuition, or small-scale pilots that are slow and costly to run. This paper presents Meta Society, a smart- governance platform that lets a policymaker state a goal in natural language, receive structured policy recommendations, and evaluate candidate policies against a virtual model of a town before committing to a real-world deployment. The platform combines three components. First, an AI Policy Advisor executes a multi-stage large-language-model (LLM) reasoning pipeline-orchestrated with LangGraph and served through a FastAPI backend-that turns a natural-language goal and multi-domain town statistics into structured candidate policies. Second, a Simulation Lab reports its outcomes from a deterministic engine rather than from an LLM: one agent per citizen is generated for a real town (Pandharpur, Solapur district, Maharashtra) so that published Census 2011 totals are reproduced exactly; a 35-node Bayesian network- thirty core nodes plus five exogenous shock roots-is rolled forward one period at a time with conditional-probability tables counted from the population wherever the data can speak; a historical-grounding stage transfers characteristics from real predecessor programmes to the proposed policy and derives a grounded effect scale that the causal network consumes; an external-shock layer stress-tests the policy under hypothetical events; and a Differential Evolution search with NSGA-II non-dominated selection, reported alongside an equal-budget random-search control, explores the policy- parameter space. Third, a data and dashboard layer renders the same population together with a field-by-field provenance ledger that separates verified Census fields from modelled ones. Every headline metric is a stated summation over the agent population, and each run re-checks the accounting identities, the network’s documented response directions, and reproducibility, reporting the outcome as an evidence pack. We describe the architecture, the modules, the workflow, and the verification the engine performs, report a seed-ensemble uncertainty protocol and a draw-count sensitivity analysis, and state the system’s boundaries plainly: response directions are validated but magnitudes are not calibrated against an evaluated real programme, income and sector are modelled rather than measured, and the platform is a counterfactual comparison tool rather than a forecasting service.
Keywords: smart governance, policy simulation, large language models, LangGraph, Bayesian networks, differential evolution, NSGA-II, agent-based modeling, microsimulation, synthetic population, external shocks, historical grounding.
Keywords: smart governance, policy simulation, large language models, LangGraph, Bayesian networks, differential evolution, NSGA-II, agent-based modeling, microsimulation, synthetic population, external shocks, historical grounding.
Abstract
Smart AI-Powered E-Commerce Platform with Personalized Recommendation and Sales Analytics
Wadekar Gargi Prashant, Bole Ishwari Santosh, Patil Harshvardhan Sunil, Mr. S. S. Ingole
DOI: 10.17148/IJARCCE.2026.151046
Abstract: Online retail has become a primary shopping channel, and customers increasingly expect experiences that are tailored to them [1]. Many small and mid-sized merchants, however, still operate storefronts that show every visitor the same catalog and give owners little insight into what is selling, what is running out, and why. This paper presents a Smart AI-Powered E-Commerce Platform that combines personalized product recommendation, a database-grounded multilingual shopping assistant, and an analytics-driven administration console in a single full-stack web application. The recommendation module ranks products with an explainable score built from category affinity, popularity, customer rating, and price suitability, and derives “frequently bought together” suggestions from real co-purchase history. The shopping assistant interprets free-text queries in English, Hindi, and Marathi, retrieves products only from the store database, and phrases its replies through a language-model layer with an automatic rule-based fallback, so it never invents items or prices. For the merchant, the platform provides sales-trend, category, and regional analytics together with an inventory module that estimates sales velocity and proposes restock quantities. The system is implemented with React, Node.js/Express, and MongoDB, and includes JWT-based sessions, e-mail one-time-password two-factor login, server- side price computation, Razorpay online payment, and cash on delivery. A working prototype comprising 89 REST endpoints and 12 data collections demonstrates the design across customer and administrator interfaces. The platform is expected to improve product discovery, reduce stock-outs, and support data-driven decisions for local retailers; future work covers collaborative filtering, payment-gateway webhooks, and large-scale usability and load evaluation.
Keywords: E-Commerce, Personalized Recommendation, Sales Analytics, AI Shopping Assistant, Inventory Intelligence, MERN Stack
Keywords: E-Commerce, Personalized Recommendation, Sales Analytics, AI Shopping Assistant, Inventory Intelligence, MERN Stack
Abstract
IOT-Based Machine Health Monitoring System
Chavan Rutuja Umesh, Babar Dnyaneshwari Namdev, Gajare Pragati Dnyaneshwar, Mrs. P. S. Gavali
DOI: 10.17148/IJARCCE.2026.151047
Abstract: Motors, pumps, compressors, generators and conveyors underpin modern manufacturing, but each can break down without warning, leading to unplanned stoppages, expensive repairs and growing maintenance costs. This paper describes an Internet of Things (IOT) system that continuously supervises such equipment. An ESP32 microcontroller reads temperature and vibration sensors and transmits the readings over Wi-Fi to a FastAPI backend, which stores them in a MySQL database and serves them to a ReactJS dashboard. Each reading is evaluated against limits configured for the individual machine, and the machine is classified as Healthy, Warning or Critical. A Critical state triggers a local buzzer and a remote notification through Blynk Cloud. In addition, a machine learning module examines the stored sensor history to anticipate probable faults and so supports predictive maintenance. The result is a low-cost, intelligent approach to industrial condition monitoring.
Keywords: IOT, Predictive Maintenance, Machine Health Monitoring, Machine Learning, ESP32
Keywords: IOT, Predictive Maintenance, Machine Health Monitoring, Machine Learning, ESP32
Abstract
AI-Based Road Accident Detection and Alert System
Mr. S. A. Zambare, Saniya Mainuddin Mulla, Kiran Sanjay Jadhav, Shivprasad Suresh Kadam, Reshma Mujofar Pathan
DOI: 10.17148/IJARCCE.2026.151048
Abstract: Road accidents cause heavy loss of life, and the outcome for victims depends strongly on how quickly emergency services are informed. Manual CCTV monitoring is slow and error-prone. This paper presents the design, architecture and implementation of an AI-based Road Accident Detection and Alert System that analyses CCTV or user- uploaded video frame by frame using Open CV and a YOLO deep-learning detector, and forwards confirmed accidents to emergency services. The system is built with Python, Flask, HTML, CSS and JavaScript and offers a public accident- reporting page with automatic GPS location capture, an automatic video-processing pipeline, an accident review dashboard with statistics, and a separate police/ambulance operations console with live alerts and filters. In the test session the dashboard logged 42 detected events, of which 16 were confirmed on review (38.1 %). The work shows that automated detection with human confirmation and location-tagged alerts can shorten the path from an accident to emergency response, and it identifies precision improvement as the main area for further work.
Keywords: Road Accident Detection, Computer Vision, Deep Learning, YOLO, Open CV, CCTV Surveillance, Emergency Alert System, Django, Intelligent Transportation System.
Keywords: Road Accident Detection, Computer Vision, Deep Learning, YOLO, Open CV, CCTV Surveillance, Emergency Alert System, Django, Intelligent Transportation System.
Abstract
Smart Village Automation Using IoT
Ms. J. K. Hipparkar, Pooja Karande, Shreya Joshi, Arjiya Mulani
DOI: 10.17148/IJARCCE.2026.151049
Abstract: Water is one of the most crucial resources of a village, but its availability and demand is majorly calculated manually, thereby leading to delayed decision-making and poor planning. Smart Village Automation Using IoT is a low- cost solution which leverages the power of the Internet Of Things and Machine Learning to manage the water distribution of rural areas and village services [1] [2]. An ultrasonic sensor is used to calculate the water level of the storage tank, which is then transmitted via Wi-Fi (MQTT) to an application server which validates the data and stores it in a MongoDB database. The Python FastAPI server is used to host the machine learning models; a demand prediction model that predicts the demand based on water level, temperature, rainfall, population, date-time and water used, and a Random Forest Regressor model which classifies between normal consumption and leakage/theft. The citizen and moderator portals built using React allow users to view water-related information, government schemes, complaints, and Razorpay payment integration. The complaint management system escalates unaddressed complaints to the relevant authority after 25 days of creation. In the leakage/theft model, the pressure has a feature importance of 48.3%, followed by the actual consumption at 26.8%, flow rate at 18.1%, and expected consumption at 6.8%. The low-cost implementation has only one hardware sensor, the ultrasonic sensor, but factors like high implementation cost, lack of technical know-how, connectivity issues, training data, and maintenance can affect the adoption rate. Future improvements include the addition of more sensors and tanks, better training data, deploying the application on mobile and cloud platforms and exploring other avenues of Smart Village such as agricultural support and waste management.
Keywords: IoT, Smart Village, Machine Learning, Water Distribution, Water Demand Prediction, Leakage Detection, ESP32, Ultrasonic Sensor
Keywords: IoT, Smart Village, Machine Learning, Water Distribution, Water Demand Prediction, Leakage Detection, ESP32, Ultrasonic Sensor
Abstract
VocaLyze – AI Based Emotion Recognition from Voice
Sadigale Shravani Santosh, Mali Vaishnavi Sukhdev, Gotavle Sushma Sanjay, Linge Sandeep Ganpat
DOI: 10.17148/IJARCCE.2026.151050
Abstract: When a Person emotion change Changes can be seen in their voice tone, pitch speaking speed and intensity. The name of our project is VocaLyze. It is an AI-based web application that identifies a person emotion from their voice. In this system the user can provide their voice in two ways first by recording their voice using a microphone and second by uploading an existing audio file. After receiving the voice, the system first removes background noise and unwanted silent parts from the audio. Then the system extracts voice features such as MFCC, pitch, energy and spectral features. These features are then given to machine Learning and Deep Learning models such as Random Forest, SVM, CNN and LSTM. These models help the system to identify and classify the emotion present in the user voice. The system also provides a confidence score for the detected emotion. Our project does not only identify the emotion. It also provides recommendations according to the detected emotion. These recommendations can include quotes, calming music, breathing, meditation and wellness activities.
Keywords: Artificial Intelligence, Machine Learning, Voice Emotion Recognition, MFCC, CNN, LSTM, Emotion Classification, Personalized Recommendation.
Keywords: Artificial Intelligence, Machine Learning, Voice Emotion Recognition, MFCC, CNN, LSTM, Emotion Classification, Personalized Recommendation.
Abstract
An Adaptive and Risk-Aware Access Control Framework for Secure Cloud File Sharing Using Hybrid Cryptography and Context-Aware Key Release
Ghodekar Snehal Ramesh, Ghodekar Vaishnavi Suresh, Miss. Ghodake G. K.
DOI: 10.17148/IJARCCE.2026.151051
Abstract: Cloud platforms have become a common place to keep business records because they are easy to use and accessible from anywhere. Still, sensitive documents stored this way face two problems: the service provider should not be able to read them, and a password alone should not decide whether a request is genuine. This study proposes ARAC- HC, a framework where the key needed to open a document is released only after the system judges the situation of each request. At upload time, the document is protected on the user’s own device with a 256-bit AES-GCM key. This key is then wrapped separately for every approved reader through a short-lived X25519 exchange followed by HKDF. Before handing over the wrapped key, the framework checks seven signals: device state, location, access time, failed login attempts, recent user activity, stored trust level and the sensitivity of the requested document. Based on the calculated risk, the system either allows normal access, asks for an extra check, permits restricted access or refuses the request. In tests, protecting a 100 MB document took close to 75 ms. The X25519-based method finished key protection roughly 2.6 times faster than RSA-2048. Across 20,000 simulated requests, the framework caught 95.3% of malicious attempts, and 97% of legitimate users got access without any extra verification step.
Keywords: Context-aware key release, risk-based authorisation, cloud document sharing, envelope encryption, AES- GCM, X25519, insider misuse, zero trust.
Keywords: Context-aware key release, risk-based authorisation, cloud document sharing, envelope encryption, AES- GCM, X25519, insider misuse, zero trust.
Abstract
Information Security Copilot for Cyber Threat Monitoring and Risk Management in Small and Medium-Sized Businesses
Yashowardhan Thakare, Rutuja Ghongade, Payal Pawar, Radha Mankar, Anisha Kaware, Dr V B Gadicha
DOI: 10.17148/IJARCCE.2026.151052
Abstract: Small and Medium Businesses (SMBs) increasingly depend on cloud platforms, email, CRM systems, and other digital services, while limited budgets and cybersecurity expertise increase their exposure to threats. This paper proposes an AI Security Copilot for continuous security monitoring in SMB environments. The system collects security data through authorized APIs, analyzes activities and anomalies using AI/ML techniques, prioritizes risks, and provides actionable security recommendations through a centralized dashboard. The proposed approach aims to make cybersecurity monitoring more accessible, manageable, and effective for resource-constrained
Keywords: AI Security Copilot, Cybersecurity, SMB Security, Threat Detection, Anomaly Detection, Risk Assessment, Security Monitoring.
Keywords: AI Security Copilot, Cybersecurity, SMB Security, Threat Detection, Anomaly Detection, Risk Assessment, Security Monitoring.
Abstract
Smart Public Infrastructure Damage Reporting and Prediction System: A Web-Based Platform with Category-Based Rule Priority Classification
Khushi, Harsha Bhat T.M, Varun P, Shreecharan I G, Prof. Raksha K
DOI: 10.17148/IJARCCE.2026.151053
Abstract: Potholes, damaged roads, faulty streetlights and water leakage often remain unresolved because citizens lack a simple reporting channel and authorities lack consolidated data. This paper presents a web-based Smart Public Infrastructure Damage Reporting and Prediction System built with HTML, CSS, JavaScript, PHP and MySQL. Citizens register with OTP verification, submit geotagged complaints with photographs, and track status; administrators manage status, priority and department assignment, and view complaints on an interactive map and analytics dashboard. Source- code tracing shows priority is assigned by a transparent category rule (High for Water Leakage, Pothole and Traffic Signal Failure, otherwise Medium), while several interface features labeled “AI” are fixed values or dormant code. Functional testing of ten cases passed. The paper documents the system as built and outlines steps to close this gap.
Keywords: infrastructure complaint reporting, civic technology, rule-based prioritization, map visualization, web application, PHP, MySQL.
Keywords: infrastructure complaint reporting, civic technology, rule-based prioritization, map visualization, web application, PHP, MySQL.
Abstract
AI in Healthcare Applications: A Study of Patient Awareness and Acceptance
Priyanka Ambdekar, Aditi Hingane, Prof. Pramod Gopale
DOI: 10.17148/IJARCCE.2026.151055
Abstract: Artificial Intelligence (AI) is becoming an important part of the healthcare sector by providing innovative solutions to improve patient care and healthcare services. However, the increasing demand for medical services and the shortage of healthcare professionals have created a need for advanced technologies in healthcare. AI-based applications, such as healthcare chatbots, care robots, mental health support systems, and machine learning-based diagnostic tools, are being developed to assist patients and healthcare professionals. These technologies can help improve healthcare accessibility, support early disease detection, and provide better patient assistance. Despite these benefits, the successful implementation of AI in healthcare depends on patients' awareness, trust, and acceptance of these technologies. Factors such as privacy concerns, reliability, safety, and a lack of knowledge may influence patients' willingness to use AI-based healthcare applications. Therefore, this study focuses on understanding patient awareness and acceptance of AI-based healthcare applications and identifying the factors that influence their adoption. The findings may help in understanding patients' perceptions of AI in healthcare and encourage the effective use of AI technologies in medical services.
Keywords: Artificial Intelligence (AI), AI in Healthcare, Healthcare Applications, Patient Awareness, Patient Acceptance, Patient Perception, Patient Trust, Machine Learning, Healthcare Technology, Data Privacy, Technology Adoption.
Keywords: Artificial Intelligence (AI), AI in Healthcare, Healthcare Applications, Patient Awareness, Patient Acceptance, Patient Perception, Patient Trust, Machine Learning, Healthcare Technology, Data Privacy, Technology Adoption.
Abstract
CNN-Based Framework for Maize leaf Disease Detection and Crop Monitoring
Dr.K.L.S.Soujanya, V. Sumedha, M. R. V. Sravanthi, A. Supriya
DOI: 10.17148/IJARCCE.2026.151056
Abstract: Maize is an agricultural crop but its productivity is significantly affected by diseases that are often hard to notice at an early stage. Traditional ways of identifying diseases depend on people looking at the plants, which can take a lot of time and needs someone who knows a lot about plants. This paper shows a Smart Maize Crop Monitoring and Disease Detection System that uses machine learning and image processing to help with identifying and managing diseases on maize leaves. The system uses a Convolutional Neural Network to sort images of maize leaves into four groups: Healthy, Common Rust, Blight and Gray Leaf Spot. The images are changed in size. Made different in ways to help the model work better when the pictures are in different positions or under different conditions. The model that was trained is part of a web app built with Flask so people can send images of maize leaves. Get predictions about the disease through an easy way to use. Also the system gives information about how to treat the disease advice on what stage the crop's in and tips on watering the plants. Testing showed that the system could correctly say what the disease was 92% of the time which shows that this way of doing things can be a tool, for growing maize. The system is meant to help find diseases stop people from using the wrong treatments and make growing crops better.
Keywords: Maize Disease Detection, Convolutional Neural Network (CNN), Image Processing, Machine Learning, Smart Agriculture, Flask Web Application.
Keywords: Maize Disease Detection, Convolutional Neural Network (CNN), Image Processing, Machine Learning, Smart Agriculture, Flask Web Application.
Abstract
A ROBUST FRAMEWORK FOR COPYWRIGHT PROTECTION AND OWNERSHIP VERIFICATION OF DIGITAL IMAGES
Pavithra N, Anushree P M, Kavana K, Megha Manjappa Mogali, V Tejaswini
DOI: 10.17148/IJARCCE.2026.151057
Abstract: The widespread sharing of digital images through websites and social media has made unauthorized copying, redistribution, and manipulation easier. These issues create difficulties for photographers, designers, content creators, and organizations seeking to establish ownership of their work. This paper presents a robust framework for copyright protection and ownership verification of digital images using digital watermarking and cryptographic hashing. The proposed system is designed to register an image with creator metadata, embed copyright information as a watermark, generate a SHA-256 hash, and store the related ownership record for later verification. During verification, the submitted image is checked against stored records; a hash mismatch indicates that the file differs from the registered version, while watermark extraction can provide an additional ownership check when supported by the implemented watermarking method. The framework is planned around Python and image- processing libraries, with a database for ownership metadata. The design aims to combine image-level copyright marking with cryptographic integrity checking. Quantitative performance and robustness must be established through implementation and testing; this paper therefore does not claim unmeasured accuracy or performance results.
Keywords: Digital image copyright, digital watermarking, SHA-256, cryptographic hashing, ownership verification, image integrity, Python.
Keywords: Digital image copyright, digital watermarking, SHA-256, cryptographic hashing, ownership verification, image integrity, Python.
Abstract
An Intelligent Machine Learning Framework for Early Cardiovascular Disease Risk Prediction
Gnanamani. H, Gagana G, Inchara Bhatta K.M, Kavana, Likitha. H
DOI: 10.17148/IJARCCE.2026.151058
Abstract: Heart disease is one of the leading causes of death worldwide, making early prediction essential for improving patient outcomes. Machine learning techniques have gained significant attention for predicting heart disease by analyzing medical and clinical data. This survey paper reviews recent studies on heart disease prediction using various machine learning algorithms, including Logistic Regression, Support Vector Machine, Decision Tree, Random Forest, K-Nearest Neighbors, Gradient Boosting, and XGBoost. It also discusses feature selection techniques, data preprocessing methods such as SMOTE, and model optimization approaches that improve prediction performance. Based on the reviewed literature, ensemble learning methods, particularly XGBoost and Gradient Boosting, demonstrate superior accuracy and reliability. This survey highlights the strengths, limitations, and future scope of machine learning-based heart disease prediction systems and emphasizes their potential to support early diagnosis and clinical decision-making.
Keywords: Heart Disease Prediction, Machine Learning, Cardiovascular Disease, Classification Algorithms, Explainable Artificial Intelligence (XAI), Feature Selection, SMOTE, Healthcare Analytics.
Keywords: Heart Disease Prediction, Machine Learning, Cardiovascular Disease, Classification Algorithms, Explainable Artificial Intelligence (XAI), Feature Selection, SMOTE, Healthcare Analytics.
Abstract
DETECTING AND MITIGATING HALLUCINATIONS IN LARGE LANGUAGE MODELS: A REVIEW OF UNCERTAINTY, EVIDENCE GROUNDING, AND AGENTIC WORKFLOWS
Anindya Singh, Rishiraj Singh
DOI: 10.17148/IJARCCE.2026.151059
Abstract: Large language models (LLMs) can give answers that sound smooth and believable but may still be wrong, unsupported, or self-contradictory. This becomes a bigger concern when LLMs are used in retrieval systems and agent- based workflows, where they may choose tools, understand user needs, search databases, and perform actions. In these situations, even one incorrect claim can affect later decisions.
This review looks at research on LLM truthfulness, hallucination detection, uncertainty, response consistency, retrieval- augmented generation (RAG), internal model signals, and agent reliability. It groups existing methods based on the evidence they use agreement between generated answers, prediction uncertainty, model-internal information, external retrieved sources, and workflow context.
One key point is that no single signal is enough to detect hallucinations. A model can be confident but wrong, retrieved information can be incomplete or misleading, and several answers may agree without being correct.
Therefore, reducing hallucinations requires combining different types of evidence.
To address this gap, the paper proposes a unified evidence-integration framework. The framework combines token-level probabilities and entropy, hidden-state and attention-based features where available, sensitivity to prompt changes, the quality and relevance of retrieved sources, claim-level support, and domain-specific context. In addition, it incorporates a confidence-score check and intent analysis using a classifier model to improve the reliability of the overall decision process.
Using this risk estimate, the system can answer directly, verify the response again, retrieve more information, ask the user for clarification, or avoid answering when confidence is low. The Model Context Protocol (MCP) is discussed to link agents with external tools and data, but it does not by itself ensure that the information is accurate. Because the project materials describe a proposed method rather than completed experiments, this review does not include new benchmark results. It ends by highlighting future evaluation areas such as detection accuracy, calibration, robustness, speed, cost, and safe action execution in agentic systems.
Keywords: large language models; hallucination detection; factuality; semantic entropy; uncertainty estimation; retrieval-augmented generation; Model Context Protocol; agentic AI; selective prediction
This review looks at research on LLM truthfulness, hallucination detection, uncertainty, response consistency, retrieval- augmented generation (RAG), internal model signals, and agent reliability. It groups existing methods based on the evidence they use agreement between generated answers, prediction uncertainty, model-internal information, external retrieved sources, and workflow context.
One key point is that no single signal is enough to detect hallucinations. A model can be confident but wrong, retrieved information can be incomplete or misleading, and several answers may agree without being correct.
Therefore, reducing hallucinations requires combining different types of evidence.
To address this gap, the paper proposes a unified evidence-integration framework. The framework combines token-level probabilities and entropy, hidden-state and attention-based features where available, sensitivity to prompt changes, the quality and relevance of retrieved sources, claim-level support, and domain-specific context. In addition, it incorporates a confidence-score check and intent analysis using a classifier model to improve the reliability of the overall decision process.
Using this risk estimate, the system can answer directly, verify the response again, retrieve more information, ask the user for clarification, or avoid answering when confidence is low. The Model Context Protocol (MCP) is discussed to link agents with external tools and data, but it does not by itself ensure that the information is accurate. Because the project materials describe a proposed method rather than completed experiments, this review does not include new benchmark results. It ends by highlighting future evaluation areas such as detection accuracy, calibration, robustness, speed, cost, and safe action execution in agentic systems.
Keywords: large language models; hallucination detection; factuality; semantic entropy; uncertainty estimation; retrieval-augmented generation; Model Context Protocol; agentic AI; selective prediction
Abstract
Blockchain Based Academic Certificate Verification System
Rutuja Dudhate, Shubham Gadge, Pramod Gopale
DOI: 10.17148/IJARCCE.2026.151060
Abstract: In most organizations, identity is still checked against a central database. This is convenient to run. The catch is that one party ends up owning the records, and people must keep uploading the same documents every time they sign up for something new. In this paper we propose a hybrid framework for verifying academic certificates that uses a blockchain for one job only: keeping a shared record of trust. Each issuer and holder gets a decentralized identifier (DID), and the claims go inside verifiable credentials (VCs). For every credential, a smart contract stores just its hash and its status, so nothing personal is written on the chain. We describe the architecture, follow a credential from issue to revocation, give an eight-step verification algorithm, and discuss the threats and a test plan. We have not built the prototype yet, so no performance numbers appear here; we will report them after we have measured them.
Keywords: Blockchain, decentralized identifiers, digital identity, privacy, self-sovereign identity, verifiable credentials
Keywords: Blockchain, decentralized identifiers, digital identity, privacy, self-sovereign identity, verifiable credentials
Abstract
AI-Based Fraud Detection in Digital Payment Systems
Puja Patil, Shruti Pingat, Prof. Nikita More
DOI: 10.17148/IJARCCE.2026.151061
Abstract: Digital payment systems such as Unified Payments Interface (UPI), mobile wallets, internet banking and card-based payments have made transactions faster and more convenient. At the same time, digital transactions can be targeted by card-not-present fraud, account takeover, phishing, social engineering, suspicious beneficiary activity and other forms of payment abuse. Traditional rule-based systems are useful for identifying known patterns, but fixed rules may struggle when fraud strategies change or when legitimate customer behaviour varies. This paper proposes an AI- based fraud detection framework that combines transaction-risk indicators, machine-learning classification and rule- based checks to identify potentially fraudulent digital payments. The framework considers features such as transaction amount, frequency, time, device or session changes, beneficiary history and deviations from a customer’s usual activity, subject to lawful access and privacy safeguards. A risk-scoring component classifies transactions as low, medium or high risk and recommends approval, additional verification or temporary review. The paper presents a conceptual system design and assessment methodology rather than claiming results from an experiment. Its practical value should be evaluated using representative, appropriately protected transaction data and metrics such as precision, recall, F1-score, false-positive rate and detection latency. The proposed approach aims to support timely fraud prevention while preserving customer convenience, explainability and responsible handling of financial data.
Keywords: Artificial Intelligence (AI), Digital Payments, Fraud Detection, Machine Learning, UPI, Anomaly Detection, Transaction Risk Scoring, Financial Cybersecurity, Data Privacy.
Keywords: Artificial Intelligence (AI), Digital Payments, Fraud Detection, Machine Learning, UPI, Anomaly Detection, Transaction Risk Scoring, Financial Cybersecurity, Data Privacy.
Abstract
A Review of AI-Driven Cyber Emergency Response Systems: Edge Intelligence, Digital Cyber First Aid, and Human-Centric Cybersecurity
Devendra Singh Kushwaha, Uday Payghon, Manjushree Dighe, Riya Pagar, Anushka Shinde
DOI: 10.17148/IJARCCE.2026.151062
Abstract: In the area of cyber security research, many advances have been made in the areas of automated threat detection, malware analysis, phishing identification, anomaly detection, and cyber threat intelligence. However, detection capability alone does not ensure effective assistance to the victims during active cyber incidents. This review explores the intersection of Artificial Intelligence (AI), Edge Intelligence, digital incident response, Digital Cyber First Aid, multilingual natural language processing and human-centric cyber security. The literature is organized around the cyber- incident life cycle and in particular in terms of latency, privacy, explainability, accessibility, evidence preservation, incident reporting, and oversight by humans. The review reveals a persistent fragmentation problem, i.e., many existing approaches are optimizing an individual task, while relatively few approaches connect detection, triage, immediate user guidance, structured reporting, stakeholder coordination, recovery, and post-incident learning. Based on this synthesis, an integrated reference architecture is proposed that places privacy-preserving edge triage at the user boundary and connects it to AI-assisted threat intelligence, Digital Cyber First Aid, multilingual interaction, secure case management, and authority-aware coordination. A risk-oriented evaluation framework is also defined to assess response time, triage quality, privacy, usability, accessibility, evidence integrity, and resilience. The proposed architecture is presented as a research framework rather than as an experimentally validated product; consequently, quantitative performance claims are not made. The review concludes that future cyber emergency systems should move from detection-centric automation toward trustworthy, human-in-the-loop, lifecycle-aware response ecosystems.
Keywords: Artificial intelligence, cyber emergency response, edge intelligence, digital cyber first aid, human-centric cyber security, multilingual NLP, incident response, cyber resilience, threat intelligence.
Keywords: Artificial intelligence, cyber emergency response, edge intelligence, digital cyber first aid, human-centric cyber security, multilingual NLP, incident response, cyber resilience, threat intelligence.
Abstract
AutoAssist AI: An Integrated Intelligent Platform for Vehicle Assistance and Automotive Marketplace Management
Vedant Virender Kalpande, Mohammad Huzaifa, Akhilesh Pimpalgaonkar, Abhishek Shende
DOI: 10.17148/IJARCCE.2026.151063
Abstract: AutoAssist AI is an integrated intelligent platform designed to simplify vehicle-related services through a single digital application. The system combines roadside assistance requests, AI-based vehicle price estimation, second- hand car listings, automotive spare-parts sales, and vehicle service bookings. It also provides emergency contact information to help users access essential services when required. The business dashboard allows automotive businesses to manage their vehicle listings, spare-parts products, and customer orders. The proposed platform aims to improve accessibility, convenience, and efficiency for vehicle owners, buyers, sellers, and automotive service providers. The research presents the system's major modules, proposed architecture, development methodology, and validation approach. Functional testing and usability evaluation can be used to assess the system's performance, while appropriate evaluation metrics can be applied to examine the accuracy of vehicle price estimation. AutoAssist AI provides a unified approach to vehicle assistance and automotive marketplace management.
Keywords: AutoAssist AI, Vehicle Assistance, Artificial Intelligence, Vehicle Price Estimation, Roadside Assistance, Automotive Marketplace, Second-Hand Cars, Spare Parts Management, Service Booking.
Keywords: AutoAssist AI, Vehicle Assistance, Artificial Intelligence, Vehicle Price Estimation, Roadside Assistance, Automotive Marketplace, Second-Hand Cars, Spare Parts Management, Service Booking.
Abstract
Developing an Integrated Big Data Analytics Framework for Personalized E-Learning in Open and Distance Learning Institutions in Kenya
Antony Odhiambo Okeyo*, Daniel Khaoya, Bostley Muyembe
DOI: 10.17148/IJARCCE.2026.151064
Abstract: Big Data Analytics (BDA) offers open and distance learning (ODL) institutions a means of converting learner data into personalized support, yet many institutions in Kenya lack structured guidance for implementing it. Learner data remain fragmented across systems and are rarely translated into timely personalized interventions, a situation aggravated by infrastructure, skills and governance gaps. This study aimed to develop a framework for implementing BDA in personalized e-learning environments in ODL institutions in Kenya. A case study design was applied at the Open University of Kenya. Stratified, simple random and purposive sampling produced a sample of 293 students and staff, of whom 279 (95.2%) returned questionnaires. Interviews, collaborative framework meetings and document review supplemented the survey. Data were analysed using descriptive statistics, t-tests, ANOVA, multiple regression, thematic and content analysis, and partial least squares structural equation modelling. BDA (β = 0.42) and learning analytics (β = 0.36) significantly predicted personalized e-learning, and the model explained 62% of its variance. Faculty adoption and proficiency (β = 0.18) and quality of data infrastructure (β = 0.21) significantly moderated these effects. Document review and interviews exposed fragmented systems, weak governance and capacity gaps. The resulting Integrated Big Data Analytics–Enhanced Personalized Electronic Learning (IBDA-EPEL) framework showed acceptable reliability (composite reliability 0.88–0.93; AVE 0.62–0.73) and was validated by eight experts (mean 4.75/5). The study concludes that effective personalization requires integrated data, learning analytics, competent faculty, reliable infrastructure and governance. ODL institutions should adopt the framework through phased implementation supported by data integration, staff development and data governance.
Keywords: Big Data Analytics; Learning Analytics; Personalized E-Learning; Open and Distance Learning; Framework Development; Kenya
Keywords: Big Data Analytics; Learning Analytics; Personalized E-Learning; Open and Distance Learning; Framework Development; Kenya
Abstract
A Voice-First Personal Finance Assistant Using Tool Calling and a Locally Served Fine-Tuned Small Language Model
Rishi Raj Sharma, MD Faiz, S. Madan, Tejashwini, Udaya Shankar Budagavi
DOI: 10.17148/IJARCCE.2026.151065
Abstract: Large language model assistants can answer questions about personal finances, but can also make mistakes, upload sensitive data or run actions without human approval. FIN-AI is a voice-first personal finance assistant for Indian users with local data storage, rupee amounts, loan instalments, and lakh/crore units. A model is only used to select a tool, calculations are carried out by Python, and answers are phrased with templates; model-proposed modifications are confirmed via Proceed cards. Queries follow a three-tier pipeline: offline intent classifier, a locally hosted Qwen3-1.7B with QLoRA, and cloud LLMs as fallback. Training lessons are created from the real tool registry; the first model was fine-tuned on a 4 GB laptop GPU and later ones on a free cloud T4. In 67 hand-written questions, tool-selection accuracy increased from 71.6% (untuned) to 91.0% after the first fine-tuning (exact McNemar p = 0.0002); the final version scored 89.6% (Wilson 95% interval 80.0–94.8%). The fast path handled 98 of 105 read-only questions offline. Development evaluations are presented, and the data sent to external AI, speech, and market-data services is detailed.
Keywords: Voice assistant, Personal finance, Tool calling, Small language model, QLoRA, Query routing.
Keywords: Voice assistant, Personal finance, Tool calling, Small language model, QLoRA, Query routing.
Abstract
Demand & Decision Intelligence System: An AI/ML-Based Platform for Demand Forecasting, Anomaly Detection, and Inventory Optimization in Indian Retail
Prof. L.S. Bhattad, Yash Gawande, Vaishnavi Chauhan, Qazi Ubed Rehaman, Yash Zade, Kaushal Tayde, Disha Patil
DOI: 10.17148/IJARCCE.2026.151066
Abstract: Demand forecasting and inventory management help to ensure product availability and minimize inventory costs in retail or fast-moving consumer goods (FMCG). This paper introduces the Demand & Decision Intelligence System, an artificial intelligence/machine learning-based system, which combines forecasting, inventory control, anomaly detection, price elasticity, and natural language business inquiry in one web application.
The algorithm uses open-source Indian grocery data from April 1 to July 10, 2022, comprising of about 46.7 million transactions. It evaluates eight candidate algorithms, including Naive, Seasonal Naive, Moving Average, Prophet, Ridge Regression, HistGradientBoosting, Croston’s Method, and Syntetos-Boylan Approximation (SBA). On the chronological holdout test set between July 1-10, 2022, the HistGradientBoosting algorithm had a weighted absolute percentage error (WAPE) of 30.08% and a mean absolute error (MAE) of 0.5481. Thus, the evaluated forecasting pipeline requires more testing to generalize the results.
Additionally, the application introduces an inventory decision support engine that automates safety stock, reorder point, and target stock calculations, a rolling Z-score and isolation forest anomaly detection module, log-log regression price elasticity model, and retrieval-augmented generation chatbot for business inquiries. The technical architecture consists of a React.js dashboard, FastAPI backend, and PostgreSQL database. Overall, the paper presents an integrated decision- support system that combines demand forecasting and inventory planning in a single platform. However, it requires additional research to evaluate the effectiveness of inventory suggestions, anomaly detection, price elasticity, and chatbot’s applicability.
Keywords: Demand Forecasting, Inventory Optimization, Machine Learning, Anomaly Detection, Time-Series Analysis, Business Intelligence, RAG Chatbot, Price Elasticity, Indian Retail, FMCG.
The algorithm uses open-source Indian grocery data from April 1 to July 10, 2022, comprising of about 46.7 million transactions. It evaluates eight candidate algorithms, including Naive, Seasonal Naive, Moving Average, Prophet, Ridge Regression, HistGradientBoosting, Croston’s Method, and Syntetos-Boylan Approximation (SBA). On the chronological holdout test set between July 1-10, 2022, the HistGradientBoosting algorithm had a weighted absolute percentage error (WAPE) of 30.08% and a mean absolute error (MAE) of 0.5481. Thus, the evaluated forecasting pipeline requires more testing to generalize the results.
Additionally, the application introduces an inventory decision support engine that automates safety stock, reorder point, and target stock calculations, a rolling Z-score and isolation forest anomaly detection module, log-log regression price elasticity model, and retrieval-augmented generation chatbot for business inquiries. The technical architecture consists of a React.js dashboard, FastAPI backend, and PostgreSQL database. Overall, the paper presents an integrated decision- support system that combines demand forecasting and inventory planning in a single platform. However, it requires additional research to evaluate the effectiveness of inventory suggestions, anomaly detection, price elasticity, and chatbot’s applicability.
Keywords: Demand Forecasting, Inventory Optimization, Machine Learning, Anomaly Detection, Time-Series Analysis, Business Intelligence, RAG Chatbot, Price Elasticity, Indian Retail, FMCG.
Abstract
VitaDetect: Deep Learning Based Vitamin Deficiency Detection System
Anchith Shanthivana C, Anjali V, Lakshmi Arpan, Pratham B G Rai
DOI: 10.17148/IJARCCE.2026.151067
Abstract: VitaDetect is a preliminary, non-invasive vitamin-deficiency screening prototype combining EfficientNetV2S image classification with a Multi-Layer Perceptron (MLP) trained on questionnaire data. It predicts Normal, Vitamin A, Vitamin B-12, Vitamin C, and Vitamin D classes. The image dataset was audited for duplicates, conflicting labels, and corrupted images; the symptom branch uses Scikit-Learn preprocessing and SMOTE. Since the modalities have no shared patient identifiers, recall-weighted decision-level fusion with disagreement arbitration is used. On independent test sets, image and symptom models achieved 75.61% accuracy on 783 photographs and 78.00% on 600 records. A React/Vite frontend and Flask backend support up to four photos per assessment.
Keywords: vitamin deficiency detection, deep learning, EfficientNetV2, symptom classification, multimodal learning, late fusion, uncertainty handling, data quality auditing
Keywords: vitamin deficiency detection, deep learning, EfficientNetV2, symptom classification, multimodal learning, late fusion, uncertainty handling, data quality auditing
Abstract
A Reinforcement Learning-Based Approach to Optimize Network Performance in Wireless LANs
Dr VijayaLakshmi D, Manju L, Manya V Bethur, Niriksha V Shetty, Pragathi P
DOI: 10.17148/IJARCCE.2026.151068
Abstract: The IEEE 802.11 wireless networks commonly employ the distributed coordination function (DCF) for distributed channel access among competing stations. However, the conventional contention-based backoff mechanism suffers from increased collisions, packet loss, latency, and reduced throughput under varying network loads and asymmetric traffic conditions. In this project, we present a lightweight reinforcement learning (RL)-based adaptive contention-window mechanism specifically designed to enhance the IEEE 802.11 DCF protocol. Utilizing tabular Q- learning, the proposed method maintains a Q-table to learn optimal contention-window adjustment strategies based on real-time network states. The RL agent dynamically decides to increase, decrease, or maintain the contention window, guided by a reward function that considers successful transmission rate, end-to-end latency, MAC failures, and queue pressure. Implementation and evaluation were conducted using the NS-3 network simulator across diverse traffic scenarios with varying traffic rates and transmission start times, simulating different contention levels and asymmetric traffic patterns. Comparative performance analysis against the standard IEEE 802.11 DCF and Enhanced Distributed Channel Access (EDCA) protocols demonstrates that the RL-based approach effectively improves wireless network performance under dynamic contention conditions. This project investigates a lightweight and interpretable reinforcement learning framework for adaptive medium access control in IEEE 802.11 networks.
Keywords: IEEE 802.11, DCF, EDCA, Reinforcement Learning, Tabular Q-Learning, Adaptive Contention Window, Wireless Networks
Keywords: IEEE 802.11, DCF, EDCA, Reinforcement Learning, Tabular Q-Learning, Adaptive Contention Window, Wireless Networks
Abstract
Multimodal Industrial Predictive Maintenance Using Hybrid Deep Learning and Explainable AI: A Literature Survey
Abhijith Nambiar, Ajin Ani Philip, Akshay Mahesh, Kevin T John, Merlin K Thomas, Girija V R
DOI: 10.17148/IJARCCE.2026.151069
Abstract: Predictive maintenance has become an important application of artificial intelligence for reducing unexpected equipment failures, downtime, and maintenance costs. Recent studies have investigated machine learning, deep learning, Remaining Useful Life estimation, anomaly detection, and Explainable Artificial Intelligence for improving industrial maintenance decisions. However, predictive performance alone is not sufficient for practical maintenance environments, where engineers also need to understand why a model produces a particular prediction. This literature survey reviews fifteen selected studies covering machine learning approaches, deep learning architectures, RUL prediction, anomaly detection, explainable predictive maintenance, digital twins, and maintenance servitisation. The reviewed studies show that Random Forest, Support Vector Regression, XGBoost, CNN-LSTM, LSTM, autoencoders, ensemble learning, and reinforcement learning have been explored for different predictive maintenance tasks. SHAP, LIME, L2X, and other explanation techniques have also been used to improve model transparency. The review identifies recurring challenges involving dataset availability, cross-domain generalization, preprocessing, explanation quality, real-time deployment, and integration of multiple sensor modalities. Based on these findings, the survey identifies a research direction toward multimodal industrial predictive maintenance using suitable hybrid deep learning and XAI techniques, while leaving the final model and dataset selection for subsequent experimental evaluation.
Keywords: Predictive Maintenance, Industrial Sensor Data, Deep Learning, Explainable Artificial Intelligence, SHAP, Remaining Useful Life, Industry 4.0.
Keywords: Predictive Maintenance, Industrial Sensor Data, Deep Learning, Explainable Artificial Intelligence, SHAP, Remaining Useful Life, Industry 4.0.
Abstract
AutoSecAudit: An Extensible Multi-Scanner Orchestration Framework with AI-Assisted Remediation Guidance, Precision Fuzzing, and Delta Auditing in DevSecOps
Kodi Vishruth Aithal, Thilak N, Vinaya Kumar B, Vishal Gowda B, Ranjitha R*
DOI: 10.17148/IJARCCE.2026.151070
Abstract: Automating security verification in fast-paced continuous integration and continuous deployment (CI/CD) pipelines creates operational friction: studies of developer tool adoption report that noisy output and missing actionable guidance discourage the use of automated analysis tools [13]. This paper presents AutoSecAudit, an open, extensible dynamic application security testing (DAST) orchestration framework with remediation guidance, designed for CI/CD pipelines and AI developer tooling. It combines (1) attack-surface discovery through a depth-limited recursive crawler, JavaScript route extraction, OpenAPI import, and a scope guard that checks every resolved address; (2) concurrent execution of 19 scanner plugins with heuristic false-positive reduction (baseline-response comparison, a two-phase statistical timing check, single-page-application filtering, and adaptive politeness with retry on HTTP 429/503); (3) an intelligence layer that merges duplicate findings on a canonical key, enriches CVE-tagged findings with CVSS data, maps findings to OWASP Top 10 (2021), CWE, and PCI DSS v4.0 by keyword rules, and assigns plain-English business- risk categories and a five-level posture grade; and (4) delivery features: predefined multi-framework fix recipes, fingerprint-based delta comparison between scans, severity gating through exit codes, and a Model Context Protocol (MCP) server with four tools for AI coding assistants. Applying patches is left to the assistant or developer, and closed- loop remediation is not evaluated. We evaluate the tool on two purpose-built test applications with 17 and 8 planted vulnerabilities, using 5 repetitions per setting on one Windows machine. Testbed 2 was fully detected with no false positive in all 5 runs, whereas testbed 1 was fully detected in 3 of 5 runs (mean recall 81%) because some scans reached the 300 s scan deadline. None of these measured runs reported a false positive, but one further testbed 2 run inside the test suite did (a missing-SameSite cookie finding that matched no planted item). Merging reduced raw plugin alerts by 21% and 22% on complete scans. Increasing worker threads from 1 to 16 did not shorten scans (median 53 s at k=1 and 53 s at k=16), so an earlier speedup claim is withdrawn, and 10 of 25 scans ran longer than the 56 s floor. Delta comparison was exact on two complete scans but produced spurious changes when scans were incomplete. These results come from one machine and small applications written alongside the tool; they say nothing about performance on production applications.
Keywords: Vulnerability Assessment, DevSecOps, Scanner Orchestration, Model Context Protocol (MCP), AI-Assisted Remediation, Statistical Timing Verification, Compliance Mapping, Delta Auditing, Web Security, API Security.
Keywords: Vulnerability Assessment, DevSecOps, Scanner Orchestration, Model Context Protocol (MCP), AI-Assisted Remediation, Statistical Timing Verification, Compliance Mapping, Delta Auditing, Web Security, API Security.
Abstract
SentinelML: Behavioral Machine-Learning-Based Ransomware Detection with Decoupled AI Security Analysis
Vasu Chowdary, Moulica Chowdary, D. M. Yashaswini, Rajashree M Byalal
DOI: 10.17148/IJARCCE.2026.151071
Abstract: Ransomware remains a critical threat to modern enterprise and personal computing infrastructure. Traditional signature-based detection mechanisms exhibit operational degradation when confronted with previously unobserved polymorphic or zero-day variants. This paper presents SentinelML, an autonomous endpoint detection architecture that identifies ransomware activity through 19 dynamic behavioral indicators collected entirely in user space without requiring kernel-level filter drivers. By sampling filesystem Shannon entropy, process concurrency anomalies, shadow copy deletion commands, and cryptographic module bindings, SentinelML trains a local Random Forest classifier that operates as an unprivileged host daemon. To address operational triage transparency, the local classifier is coupled with an architecturally decoupled Large Language Model (LLM) advisory explainer operating via an isolated HTTPS interface. We conduct an exhaustive technical audit of the production system: confirming that Shannon entropy accounts for 41.31% of aggregate Gini importance, documenting the operational tradeoffs of heuristic user-space proxies, analyzing platform asymmetry between Windows and Linux deployments, and evaluating the system strictly as an Intrusion Detection System (IDS). Finally, we disclose that previously reported internal metrics (0.9567 accuracy across candidate algorithms) could not be independently reproduced due to unpreserved training artifacts, and establish an open empirical evaluation protocol for reproducible validation.
Keywords: Ransomware Detection, Behavioral Analysis, Random Forest, Shannon Entropy, User-Space Monitoring, Explainable AI, Endpoint Security.
Keywords: Ransomware Detection, Behavioral Analysis, Random Forest, Shannon Entropy, User-Space Monitoring, Explainable AI, Endpoint Security.
Abstract
MAHATVAK CHIKITSA- AUTOMATED SKIN ONCOLOGY DIAGNOSIS WITH CLINICAL DECISION SUPPORT
Dhanya NG, Manoj R, Megha R, Harshitha R
DOI: 10.17148/IJARCCE.2026.151072
Abstract: Skin cancer is one of the major dermatological conditions where early identification of suspicious skin lesions can assist in timely clinical assessment. However, distinguishing different lesion types from dermoscopic images is challenging because several categories may exhibit similar visual characteristics. This project presents a deep learning- based approach for automated skin lesion classification using dermoscopic images obtained from the International Skin Imaging Collaboration (ISIC) dataset. The proposed system performs image preprocessing, resizing, normalization, augmentation, and class balancing to prepare the dataset for model development. Five transfer-learning architectures, namely VGG19, VGG16, ResNet50, MobileNet, and EfficientNet, are trained and evaluated for the classification of nine skin lesion categories. The models are compared using classification performance measures including accuracy, precision, recall, and F1-score. Among the evaluated architectures, EfficientNet achieves the highest performance with a test accuracy of 90.11% and a macro F1-score of 90.57%. The selected model is further integrated into a Flask-based web application that allows users to upload a dermoscopic image and receive the predicted lesion category, confidence level, and preliminary guidance. The developed system is intended to support computer-aided screening and decision-making, while final diagnosis and treatment decisions remain with qualified healthcare professionals.
Keywords: Skin Cancer Detection, Skin Lesion Classification, Deep Learning, Transfer Learning, ISIC Dataset, Dermoscopic Images, EfficientNet, Flask.
Keywords: Skin Cancer Detection, Skin Lesion Classification, Deep Learning, Transfer Learning, ISIC Dataset, Dermoscopic Images, EfficientNet, Flask.
Abstract
AI-Powered Smart Glasses for Digital Eye Strain and Drowsiness Detection
Sanmaya I K, Sharanya S, Vinusha D Souza, Maneesha, Bharathi K
DOI: 10.17148/IJARCCE.2026.151073
Abstract: Prolonged digital screen usage can contribute to visual discomfort, eye fatigue, and drowsiness, particularly during extended periods of computer use. This paper presents an AI-powered smart glasses system designed to monitor eye activity and identify patterns associated with digital eye strain and drowsiness. The proposed architecture incorporates a Convolutional Neural Network (CNN) for eye-state classification and a Long Short-Term Memory (LSTM) network for temporal analysis of eye-related features. Eye activity is monitored using facial and eye landmarks, from which eye aspect ratio and related features are extracted. Sequential analysis is used to distinguish normal eye activity from patterns associated with eye strain and drowsiness. The system architecture also includes an alert mechanism intended to notify the user when potentially concerning eye-activity patterns are detected. The implementation work includes CNN-based eye-image classification, LSTM-based state classification, and a real-time eye-monitoring prototype. The approach provides a foundation for wearable monitoring of eye activity and supports further integration into a compact smart glasses platform.
Keywords: Smart Glasses, Digital Eye Strain, Drowsiness Detection, Convolutional Neural Network, Long Short-Term Memory, Eye Aspect Ratio, Real-Time Monitoring.
Keywords: Smart Glasses, Digital Eye Strain, Drowsiness Detection, Convolutional Neural Network, Long Short-Term Memory, Eye Aspect Ratio, Real-Time Monitoring.
Abstract
Comparative Analysis of Deep Learning Models for Offensive Comment Detection in Code-Mixed Dravidian Languages, with a Focus on Tamil-English
Maria Sarah J, Dr. Golda Dilip
DOI: 10.17148/IJARCCE.2026.151074
Abstract: Offensive and hate speech detection on social media has become an important natural language processing problem because of the scale at which harmful user-generated content is produced and the practical impossibility of moderating it manually. The problem is considerably harder for code-mixed Tamil-English text, where Tamil words are frequently transliterated into Roman script, spelling is inconsistent, and sentence structure borrows freely from both languages, all of which limit the effectiveness of models built primarily for monolingual, formally written text. This paper presents a comparative review of fifteen system-description papers submitted to the DravidianLangTech- EACL2021 shared task on offensive language identification in Tamil, Kannada and Malayalam code-mixed text, evaluated on the Dravidian Offensive Language Dataset released for that task, with the base paper SJ_AJ@DravidianLangTech-EACL2021 used as the primary point of reference. The fifteen studies are compared on dataset, preprocessing, feature representation, model architecture, training strategy and reported evaluation metric, rather than evaluated in isolation. The review finds that classical machine learning with TF-IDF or character n-gram features, non-transformer deep learning architectures such as BiLSTM and CNN-BiLSTM hybrids, and pretrained multilingual transformers such as mBERT and XLM-RoBERTa, often strengthened with ensembling, task-adaptive pretraining, transliteration or class-imbalance handling, together form the dominant methodological landscape for this task. Reported weighted F1-scores vary considerably across the three languages and techniques, with a severe class imbalance in the underlying dataset (over 70% of Tamil-English comments labelled Not-offensive) repeatedly identified as a limiting factor. The review further finds that interpretability of model predictions is addressed only marginally, mostly through confusion-matrix error analysis, and that no reviewed study applies a dedicated explainability technique. This observation motivates the proposed direction of incorporating explainable deep learning methods into Tamil-English code-mixed hate speech detection, so that model decisions can be examined for correctness as well as for the reasoning behind them.
Keywords: Hate Speech Detection; Code-Mixed Text; Tamil-English; Offensive Language Identification; Transformer Models; Explainable AI; Comparative Study
Keywords: Hate Speech Detection; Code-Mixed Text; Tamil-English; Offensive Language Identification; Transformer Models; Explainable AI; Comparative Study
Abstract
Cyber Crime Awareness, And Fraud Prevention
Nilam Yashwant Kokanee, Asmita Bhaskar Kudnar, Prof. Sanchita Panmand
DOI: 10.17148/IJARCCE.2026.151075
Abstract: The growing use of smartphones, social media, instant messaging, digital banking, online shopping and electronic services has increased the convenience of everyday life. At the same time, members of the public may be exposed to cybercrime such as phishing, fake forwarded messages, online payment fraud, fraudulent job offers, impersonation, identity theft, fake customer-care numbers and malicious applications. This research project proposes to study cybercrime awareness and prevention practices among the general public, including students, working people, business owners and senior citizens. A descriptive survey using a structured questionnaire is proposed. Responses will be summarised using counts, percentages, tables and charts after genuine data are collected. A small Python-based awareness tool is also proposed to flag warning signs in a message and provide safety guidance. The tool will be educational only and cannot conclusively determine whether a message is fraudulent. No survey findings are claimed in this draft because actual responses have not yet been collected.
Keywords: Cyber crime, online fraud, phishing, fake messages, digital safety, cyber awareness, fraud prevention.
Keywords: Cyber crime, online fraud, phishing, fake messages, digital safety, cyber awareness, fraud prevention.
Abstract
Evaluating Privacy–Utility Trade-offs in Differentially Private Federated Learning: A Comparative Benchmarking Study
Divya Varshini M, Dr Beenarani Manoj
DOI: 10.17148/IJARCCE.2026.151077
Abstract: The rapid proliferation of mobile phones, wearables, IoT sensors, and other edge computing devices has generated vast quantities of decentralized, privacy-sensitive data, intensifying interest in Federated Learning (FL) as a distributed machine learning paradigm. By keeping raw data on local devices and sharing only model updates, FL offers an appealing alternative to centralized training. However, model updates are not inherently private: prior work has shown that shared gradients and parameters can leak sensitive information through membership inference, model inversion, and gradient reconstruction attacks. Differential Privacy (DP) has therefore emerged as the de facto standard for providing formal, quantifiable privacy guarantees in FL. In a typical Differentially Private Federated Learning (DPFL) pipeline, local model updates are clipped to bound each client’s sensitivity, and calibrated Gaussian noise is injected before (or during) server aggregation, with the cumulative privacy loss tracked by a privacy accountant.
Despite the growing body of DPFL research, the literature lacks a systematic and fair comparison of existing approaches. Individual methods are typically evaluated under different datasets, model architectures, client populations, data partitioning schemes, and privacy accounting conventions, which makes reported results difficult to compare and obscures which design choices actually drive performance. This inconsistency hinders both researchers seeking to identify promising directions and practitioners seeking to select a method for deployment.
To address this gap, we present a comprehensive comparative study of fifteen state-of-the-art DPFL methods, all evaluated on the CIFAR-10 benchmark under unified and tightly controlled experimental conditions. The selected methods span the principal design dimensions of DPFL: (i) trust models (e.g., central versus local/distributed DP), (ii) noise mechanisms, (iii) gradient and update clipping strategies, (iv) privacy accounting methods, and (v) utility- enhancement techniques aimed at mitigating the accuracy loss caused by clipping and noise. All experiments use a ResNet-18 backbone with 500 federated clients and consider both IID and Non-IID data distributions, the latter simulated via Dirichlet partitioning with concentration parameters α ∈ {0.3, 0.6} to capture varying degrees of statistical heterogeneity. Each method is evaluated across privacy budgets ε ∈ {4, 6, 8, 10}, covering a practical range from relatively strong to moderate privacy protection.
Our results show that DP-FedSAM and its sparsified variant DP-FedSAM-topk consistently outperform all other baselines across the evaluated settings. Their advantage is most pronounced under heterogeneous (Non-IID) data, where sharpness-aware optimization yields flatter local minima and updates that are more robust to clipping and noise. For example, at ε = 8 under Non-IID conditions, DP-FedSAM-topk reaches a test accuracy of 53.64%, compared with 46.74% for the DP-FedAvg baseline, an absolute improvement of approximately 7 percentage points. Beyond these headline numbers, our analysis examines how the privacy–utility trade-off varies with the privacy budget, the degree of data heterogeneity, and the choice of clipping, noise, and accounting strategies.
This study makes four main contributions: (1) a unified, reproducible evaluation framework for benchmarking DPFL methods under identical conditions; (2) a consolidated comparison table covering all fifteen methods across every data distribution and privacy budget; (3) four comparative figures that visualize the privacy–utility trade-off and robustness to data heterogeneity; and (4) actionable insights into the strengths, limitations, and open challenges of existing DPFL approaches, intended to guide both future research and practical deployment.
Keywords: Federated Learning, Differential Privacy, Privacy–Utility Trade-off, Non-IID Data, Gradient Clipping, Sharpness-Aware Minimization, Comparative Study
Despite the growing body of DPFL research, the literature lacks a systematic and fair comparison of existing approaches. Individual methods are typically evaluated under different datasets, model architectures, client populations, data partitioning schemes, and privacy accounting conventions, which makes reported results difficult to compare and obscures which design choices actually drive performance. This inconsistency hinders both researchers seeking to identify promising directions and practitioners seeking to select a method for deployment.
To address this gap, we present a comprehensive comparative study of fifteen state-of-the-art DPFL methods, all evaluated on the CIFAR-10 benchmark under unified and tightly controlled experimental conditions. The selected methods span the principal design dimensions of DPFL: (i) trust models (e.g., central versus local/distributed DP), (ii) noise mechanisms, (iii) gradient and update clipping strategies, (iv) privacy accounting methods, and (v) utility- enhancement techniques aimed at mitigating the accuracy loss caused by clipping and noise. All experiments use a ResNet-18 backbone with 500 federated clients and consider both IID and Non-IID data distributions, the latter simulated via Dirichlet partitioning with concentration parameters α ∈ {0.3, 0.6} to capture varying degrees of statistical heterogeneity. Each method is evaluated across privacy budgets ε ∈ {4, 6, 8, 10}, covering a practical range from relatively strong to moderate privacy protection.
Our results show that DP-FedSAM and its sparsified variant DP-FedSAM-topk consistently outperform all other baselines across the evaluated settings. Their advantage is most pronounced under heterogeneous (Non-IID) data, where sharpness-aware optimization yields flatter local minima and updates that are more robust to clipping and noise. For example, at ε = 8 under Non-IID conditions, DP-FedSAM-topk reaches a test accuracy of 53.64%, compared with 46.74% for the DP-FedAvg baseline, an absolute improvement of approximately 7 percentage points. Beyond these headline numbers, our analysis examines how the privacy–utility trade-off varies with the privacy budget, the degree of data heterogeneity, and the choice of clipping, noise, and accounting strategies.
This study makes four main contributions: (1) a unified, reproducible evaluation framework for benchmarking DPFL methods under identical conditions; (2) a consolidated comparison table covering all fifteen methods across every data distribution and privacy budget; (3) four comparative figures that visualize the privacy–utility trade-off and robustness to data heterogeneity; and (4) actionable insights into the strengths, limitations, and open challenges of existing DPFL approaches, intended to guide both future research and practical deployment.
Keywords: Federated Learning, Differential Privacy, Privacy–Utility Trade-off, Non-IID Data, Gradient Clipping, Sharpness-Aware Minimization, Comparative Study
Abstract
Carbon FootPrint Tracking in Logistics System: AI-Based Logistics Management Platform
Bhagyashri K Kulkarni, Shashank Urs H P, Vinod, Sandeep T Rathod, Mayur Kishan Rathod
DOI: 10.17148/IJARCCE.2026.151078
Abstract: A previous survey paper by the present authors proposed a web-based Carbon Footprint Tracking System for Logistics, combining ML-based emission prediction, eco-route optimization, shipment-level tracking, and future emission forecasting, to close gaps identified across six recent studies on AI-driven logistics carbon management. This paper evaluates the subsequently completed implementation, CarbonTrack, directly against its own source code, trained models, automated test suite, and deployment configuration rather than the earlier proposal's expectations. The platform is a React/Flask/SQLAlchemy application with 113 REST endpoints, 21 database tables, and six trained regression/classification models (Random Forest and XGBoost) covering fuel consumption, CO2 emission, fuel cost, eco score, predictive maintenance, and route recommendation, integrated with live Google Maps/Routes/Places and OpenWeather APIs. Re-executing the shipped training pipeline reproduced a predictive-maintenance accuracy of 87.5% (weighted F1 = 0.871) and a route-recommendation accuracy of 91.0%, with regression MAEs of 2.02 L (fuel), 3.68 kg (CO2), and 7.26 points (eco score) on an 80/20 held-out split. Re-executing the project's own 234-test suite reproduced 227 passing tests (97.0%). The paper discloses that ML training targets are generated by deterministic formulas from the same input features rather than independent real-world telemetry, and that the previously proposed emission-forecasting feature was not found implemented, so that measured results are not overstated as field-validated performance.
Keywords: Carbon Footprint Tracking, Logistics, Machine Learning, Random Forest, XGBoost, Predictive Maintenance, Route Recommendation, Fleet Management, Experimental Evaluation
Keywords: Carbon Footprint Tracking, Logistics, Machine Learning, Random Forest, XGBoost, Predictive Maintenance, Route Recommendation, Fleet Management, Experimental Evaluation
Abstract
Phishing Website Detection using Machine Learning
Prof. More N.N., Wayal Akanksha Sandip, Walunj Sneha Bhausaheb
DOI: 10.17148/IJARCCE.2026.151079
Abstract: Phishing is a cyber-attack in which criminals create fake websites that imitate trusted banks, online stores, and government portals to steal passwords, card numbers, and other sensitive information. Blacklist-based protection reacts only after a phishing site has been reported, while new phishing sites appear every day and many remain online for only a few hours. This research proposes a Phishing Website Detection System using Machine Learning that classifies a website as phishing or legitimate by analysing its URL, domain, and page-content characteristics. The proposed system collects a labelled dataset, extracts features such as URL length, use of an IP address, the “@” symbol, number of sub- domains, HTTPS status, domain age, and the ratio of external links, and trains supervised classifiers including Logistic Regression, Support Vector Machine, Random Forest, and XGBoost. The models are compared using accuracy, precision, recall, F1-score, and the confusion matrix, and the best model is placed behind a web application in which a user enters a link and receives a verdict with the main reasons behind it. The system supports, and does not replace, user caution, and it is designed to protect user privacy and keep false alarms on genuine websites low. Overall, the system provides a data-driven framework for early and transparent detection of phishing websites.
Abstract
Explainable AI-Based Early Dropout Risk Detection and Intervention System for Rural Schools
Pranali Santosh Wagh, Disha Santosh Waghmare, Miss. More N. N.
DOI: 10.17148/IJARCCE.2026.151080
Abstract: Student dropout is a significant challenge in rural education, often associated with irregular attendance, poor academic performance, financial difficulties, and limited access to educational resources. This research proposes an Explainable AI-Based Early Dropout Risk Detection and Intervention System for Rural Schools to identify students who may be at risk of leaving school before completing their education. The proposed system analyses relevant student data, including attendance records, academic performance, and socio-economic indicators, to estimate dropout risk using machine learning techniques. Explainable Artificial Intelligence (XAI) is incorporated to provide understandable reasons behind each prediction, enabling teachers and school administrators to make informed decisions. Based on the identified risk factors, the system can support timely interventions such as counselling, remedial education, parental engagement, and referrals to available support schemes. The proposed approach aims to improve early identification, promote transparent decision-making, and support student retention in rural schools while maintaining data privacy and fairness. Overall, the system provides a data-driven framework for proactive educational support and more inclusive learning opportunities.
Abstract
The Price of Credit Risk in Nigeria: Do Bank Lending Rates Reflect the Opportunity Cost of Treasury Investment? Evidence from Lending–Treasury Spreads, 2006-2026
Ofierohor Ufuoma Earnest
DOI: 10.17148/IJARCCE.2026.151081
Abstract: A Nigerian bank that lends to a company gives up the option of buying a Treasury bill instead. The difference between the two rates is the price at which the bank is willing to take on credit risk, illiquidity and the cost of administering a loan. This paper studies that difference using monthly Central Bank of Nigeria data for January 2006 to March 2026 (243 months). The prime lending rate exceeded the Treasury bill rate by an average of 7.29 percentage points (HAC 95% confidence interval 5.73 to 8.85), and the spread was positive in 96 per cent of months. It was not stable. It averaged 8.4 points in 2006–2014 and 9.1 points in 2021–2023, then collapsed to 1.3 points in 2024–2026, turning negative for eight consecutive months after the bill rate jumped in February 2024. Formal break tests confirm three shifts in the mean spread, in December 2008, March 2011 and November 2023. Over the same period the gap between the maximum and prime lending rates, a rough measure of the premium charged to riskier borrowers, tripled from about 5 to 15 points. The spread is highly persistent, with a half-life of about nine months, and neither inflation nor the policy rate explains its movements. Recent volatility in bill yields has a weakly negative association with the spread (p = 0.058). Out of sample, from 2021 to 2026, no model beats a no-change forecast. The paper also shows that two zero entries in the published bill- rate series create a spurious and highly significant volatility effect if they are not corrected. The results suggest that Nigerian banks priced their best borrowers increasingly close to the risk-free rate after 2024, while charging much more to everyone else.
Keywords: credit risk; interest rate spreads; opportunity cost; bank lending; Treasury bills; Nigeria
JEL classification: G21; G12; E43; C22
Keywords: credit risk; interest rate spreads; opportunity cost; bank lending; Treasury bills; Nigeria
JEL classification: G21; G12; E43; C22
Abstract
Context-Aware Conversational AI for Intelligent Travel Planning and Smart Tourism
Arathi H L, Aishwarya A U, Deekshitha H R, Harshitha S, Mahesh Hanchanl
DOI: 10.17148/IJARCCE.2026.151082
Abstract: The increasing use of digital technologies in tourism has created a demand for intelligent travel assistance systems that provide personalised recommendations, contextual information, and convenient trip planning. This paper presents an AI-powered multilingual travel guide designed to support tourism across Karnataka, India. The proposed system integrates natural language processing, machine-learning-based intent recognition, tourism information retrieval, destination recommendation, and budget-aware itinerary generation within a web-based platform. Developed using Python Flask, the application supports information about 62 tourist destinations and enables users to interact in English, Kannada, and Hindi. Its conversational engine combines Term Frequency–Inverse Document Frequency (TF-IDF)- based intent classification, keyword rules, context-aware dialogue processing, and hybrid information retrieval to generate relevant travel responses. A content-based recommendation component considers traveller interests, destination characteristics, ratings, popularity, geographical proximity, seasonality, and saved places. The itinerary planner generates day-wise travel schedules, estimates expenses, organises destinations geographically, and supports planning according to budget constraints. Voice interaction, destination exploration, user accounts, saved destinations, and itinerary export further extend the platform's functionality. The system uses SQLite by default and supports optional integration with a large language model for response generation. This paper presents the system architecture, implementation methodology, major functional modules, and approach to functional evaluation. The work demonstrates an integrated software architecture for multilingual, AI-assisted regional tourism and identifies opportunities for further evaluation and real-time tourism information integration.
Keywords: Artificial Intelligence, Smart Tourism, Conversational AI, Multilingual Chatbot, Information Retrieval, Recommendation Systems, Intelligent Trip Planning.
Keywords: Artificial Intelligence, Smart Tourism, Conversational AI, Multilingual Chatbot, Information Retrieval, Recommendation Systems, Intelligent Trip Planning.
Abstract
Hybrid Random Forest–Marine Predator Algorithm for High-Accuracy Energy Theft Detection
Devaraj V, Akila K
DOI: 10.17148/IJARCCE.2026.151083
Abstract: Electrical energy theft is an issue that persists in the energy sector and has led to financial losses and inefficiencies. The detection of electrical energy theft is often difficult with traditional approaches, which are not effective in dealing with varying consumption patterns and imbalanced datasets. The aim of this paper is to present an effective and efficient hybrid framework that uses the classification power of the Random Forest algorithm and the optimization power of the Marine Predator Algorithm. The proposed hybrid framework is effective in detecting electrical energy theft with better accuracy compared to other models. The proposed framework is significant in the sense that it offers an effective and efficient solution to the detection of electrical energy theft in the energy sector. The proposed hybrid framework is significant in the sense that it offers an effective and efficient solution to the detection of electrical energy theft in the energy sector and is an additional contribution to the field of smart grid analytics.
Abstract
A Comparative Analysis of Prediction-Based and Reconstruction-Based Deep Learning Models for Time Series Anomaly Detection in Spacecraft Telemetry
Ashwin Krishna N, Dr.H.Mary Shyni
DOI: 10.17148/IJARCCE.2026.151084
Abstract: Anomaly detection in multivariate spacecraft telemetry is usually judged with the point-adjusted (PA) F1 score, a protocol known to reward any detector that fires at least once inside a labelled segment. This paper reports a controlled comparison of three deep models, an LSTM predictor, an LSTM autoencoder and a GRU autoencoder with additive attention, on twelve channels of the NASA SMAP and MSL benchmark. All models share one preprocessing, windowing, smoothing and thresholding pipeline and are scored with both PA and strict point-wise F1; each is also set against two trivial baselines, a random detector and the norm of the raw input. Under PA the random detector reaches a micro- averaged F1 of 0.96, above every trained model (0.79–0.81), while its point-wise F1 is only 0.02 against 0.44–0.45 for the networks. The raw-signal baseline matches the networks on both protocols (0.81 and 0.46). With a fixed threshold multiplier the autoencoders score zero PA F1 on 5 channels; once the multiplier is tuned per channel the three architectures become indistinguishable within run-to-run variation, although on three SMAP channels the PA scores of every model rest on only 2–14 correctly flagged points (point-wise F1 ≤ 0.03), and two contextual-anomaly channels remain undetected by all of them. The results indicate that evaluation protocol and threshold calibration, rather than architecture, account for most of the differences reported on this benchmark.
Keyword: Spacecraft telemetry, multivariate time series, LSTM, Anomaly detection, Autoencoder, attention, SMAP, MSL, point adjustment.
Keyword: Spacecraft telemetry, multivariate time series, LSTM, Anomaly detection, Autoencoder, attention, SMAP, MSL, point adjustment.
Abstract
Beyond Contextual Embeddings: A Comparative Analysis of Feature-Engineered Classical Machine Learning and Transformer-Based Approaches for Deceptive Opinion Spam Detection
Shaliny P, Dr.S.Niveditha
DOI: 10.17148/IJARCCE.2026.151085
Abstract: We compare classical machine learning with a published fine-tuned BERT result on the Deceptive Opinion Spam Corpus (OpSpam), a public benchmark of 1,600 hotel reviews. Naive Bayes, logistic regression and linear SVM classifiers were trained on TF-IDF features under stratified 5-fold cross-validation, the protocol used by the BERT baseline we compare against. The strongest configuration joins word n-grams (1 to 3) with character n-grams (3 to 5). It reaches 91.12% accuracy on a fixed seed and 90.78% averaged over ten repeated runs, against 89.1% reported for BERT, and it trains in about two seconds per fold on a CPU. The margin needs careful reading. A plain unigram and bigram SVM already scores 89.38%, the fusion improves on the better of its two feature views by under one point, and its edge over the plain SVM is borderline (corrected resampled t-test, p = 0.053). When folds are grouped by hotel, so that test hotels never appear in training, every n-gram variant falls to about 89%, level with the BERT figure. We also audit the highest accuracy we found for this corpus, 96.03% for a RoBERTa and LSTM hybrid. It comes from an 80/20 split without cross-validation or variance estimates, and its precision, recall and accuracy are hard to reconcile with a single balanced test set. Under the BERT paper’s own protocol the n-gram fusion finishes 1.7 to 2.0 points ahead, but the lead disappears when hotels are held out.
Keywords: BERT, character n-grams, cross-validation, deceptive opinion spam, fake review detection, OpSpam, support vector machine, TF-IDF.
Keywords: BERT, character n-grams, cross-validation, deceptive opinion spam, fake review detection, OpSpam, support vector machine, TF-IDF.
Abstract
Intelligent Edge AI System for Satellite Image Analysis and Adaptive Data Transmission
Hari Krishna M.B, Siddarth.A, Niranjana G.M, Santhosh, Udai Shankar
DOI: 10.17148/IJARCCE.2026.151086
Abstract: Satellite imagery plays an important role in environmental monitoring, disaster management, and land-surface observation. However, processing large volumes of satellite images and transmitting them to ground stations can create challenges due to limited communication bandwidth, processing resources, and onboard energy availability. This paper presents an Intelligent Edge AI System for Satellite Image Analysis and Adaptive Data Transmission that integrates deep learning-based image classification with confidence-based transmission decisions. A Convolutional Neural Network (CNN) is employed to classify satellite images into four categories: Fire, Water, Cloud, and Land. Before classification, images are resized to 128 × 128 pixels, converted to RGB format, and normalized to improve input consistency. Based on the predicted class, confidence score, and simulated battery conditions, a decision engine determines whether an image should be sent, marked for compression, or ignored. The system also incorporates Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize the image regions that influence classification results. Its performance can be assessed using accuracy, precision, recall, F1-score, a confusion matrix, and prediction latency. The proposed framework combines image analysis, adaptive decision-making, and simulated energy-aware processing within a unified system. The current implementation provides a prototype for intelligent satellite image handling; further validation using reliable datasets, actual image compression, and real communication hardware is required to quantify its impact on transmission efficiency and energy consumption.
Keywords: Edge Artificial Intelligence, Satellite Image Analysis, Convolutional Neural Network, Adaptive Data Transmission, Energy-Aware Computing, Grad-CAM, Remote Sensing.
Keywords: Edge Artificial Intelligence, Satellite Image Analysis, Convolutional Neural Network, Adaptive Data Transmission, Energy-Aware Computing, Grad-CAM, Remote Sensing.
Abstract
AI-driven Renewable Energy Generation Forecasting Using Machine Learning Techniques
Arathi H L*, Sinchana C S, Partha C M, Shobha N, Shreya K S
DOI: 10.17148/IJARCCE.2026.151087
Abstract: Short-term forecasting of solar and wind generation is a prerequisite for economical plant operation, yet most reported machine-learning studies conclude at a table of error values computed on stored datasets. In earlier work the present authors reviewed five recent studies, identified six practical deficiencies, and proposed a five-layer forecasting framework; no implementation or measured result was reported at that stage. This paper reports the construction and quantitative evaluation of that framework as a working system. A web application was developed in which the browser Geolocation API and a public meteorological service supply the four environmental parameters used by the models, removing manual data entry. Four regression algorithms—linear regression, a decision tree, a random forest, and support vector regression—were trained separately for solar and wind output on a 5,922-record physics-based reference dataset and compared under identical conditions. On a 1,185-record held-out partition the selected solar model (support vector regression) attained MAE 1.340 kWh, RMSE 1.795 kWh and R² 0.9940, and the selected wind model (random forest) attained MAE 1.484 kWh, RMSE 1.932 kWh and R² 0.9939, for a reference plant rated at 100 kW per source. Linear regression proved adequate for solar output (R² 0.9930) but failed for wind (R² 0.8375); the discrepancy is traced to the piecewise-cubic turbine power curve. Residual analysis shows that the non-linear models operate at the injected noise floor of the reference data, so differences among them are not statistically meaningful, and permutation importance recovers the dominant physical driver of each source. Of the six deficiencies previously identified, three are closed, one is partially addressed, and two remain open. Because the reference dataset is synthetic, the reported accuracy characterises model capacity and pipeline correctness rather than field performance.
Keywords: Energy forecasting, machine learning, model comparison, permutation importance, random forest, renewable energy, support vector regression, weather application programming interface, web application.
Keywords: Energy forecasting, machine learning, model comparison, permutation importance, random forest, renewable energy, support vector regression, weather application programming interface, web application.
Abstract
Drift-Aware Online Ensemble Learning for Imbalanced Industrial Failure Detection: A Comparative Study
Janani M, Dr.M.Daniel Naraeshkumar
DOI: 10.17148/IJARCCE.2026.151088
Abstract: Predictive-maintenance models are usually reported on fixed, shuffled train-test splits. This hides two practical weaknesses: class imbalance, which makes accuracy misleading, and sensor drift, which static models never meet during evaluation. This paper presents a comparative study on the public AI4I 2020 dataset (10,000 records, 3.39% failures) that addresses both. First, we repeat the 80/20 evaluation protocol of a recent benchmark study. Gradient boosting with physics-informed features (machine power, temperature difference and wear-torque load) reaches a failure-class F1-score of 0.852 (precision 0.979, recall 0.754, ROC-AUC 0.979), against 0.823 published for the benchmark's tuned XGBoost model. The bootstrap 95% interval [0.774, 0.917] contains the published value, so the gain is indicative rather than conclusive. Second, we simulate sensor drift at five severities and compare static models with online learners that retrain on a sliding window, triggered by the Drift Detection Method (DDM), by a timer, or by both. The static physics-informed model falls from F1 0.768 without drift to 0.396 at severity 0.5 and 0.047 at severity 2. From severity 0.5 the best online variant has the higher point estimate, but the 95% intervals separate only from severity 1.5, where F1 is 0.313 against 0.144. Without drift the static model is ahead. Online adaptation therefore mitigates drift without curing it, because it needs timely labels and enough failure examples in each window. The study contributes a reproducible drift-aware evaluation protocol, an ablation of feature engineering, and a candid account of where adaptive learning helps and where it does not.
Keywords: predictive maintenance, concept drift, online learning, class imbalance, gradient boosting, industrial IoT, fault detection, AI4I 2020.
Keywords: predictive maintenance, concept drift, online learning, class imbalance, gradient boosting, industrial IoT, fault detection, AI4I 2020.
Abstract
A Comparative Study of Traditional and Ensemble Classifiers for Software Defect Prediction on the NASA PROMISE Repository
Sanjay I, Dr.P.Durga Devi
DOI: 10.17148/IJARCCE.2026.151089
Abstract: Software defect prediction (SDP) directs limited testing effort toward the modules most likely to fail, and the NASA PROMISE datasets are its most common benchmark. Published accuracies on these datasets differ widely, from about 76% to more than 97%, and studies rarely report the majority-class baseline or state how resampling and data splitting were organised. This paper presents a controlled comparison of ten traditional and ensemble classifiers on CM1, KC1, KC2, PC1 and JM1, using the ADASYN-based study of Kabir et al. (2025) as the reference benchmark. We vary three factors: the model family (logistic regression, k-nearest neighbours, tuned SVM, random forest, extra trees, gradient boosting, histogram gradient boosting, AdaBoost, voting and stacking), the evaluation protocol (single split, oversampling before the split, oversampling inside training folds, repeated stratified cross-validation) and the data version (original versus deduplicated). Alongside accuracy, we report balanced accuracy, sensitivity and AUC. Three findings emerge. First, 11% to 43% of modules in the original datasets duplicate another module, and removing duplicate and inconsistent rows lowers the best accuracy by up to nine points. Second, oversampling the whole dataset before splitting raises accuracy by two to eight points over a raw split, but on a test set that is half synthetic; applied inside training folds, ADASYN lowers accuracy by roughly one to five points while raising sensitivity, and the area under the ROC curve is almost unchanged. Third, model choice matters little: top-ranked models differ from logistic regression by at most 1.1 accuracy points, and on CM1 and PC1 no model clearly exceeds the majority baseline. Under repeated stratified cross-validation our best accuracy exceeds the reference study on PC1 and JM1 for both data versions, while its CM1 result is not matched. We therefore recommend reporting accuracy with the baseline, balanced accuracy and sensitivity under a leakage-free protocol.
Keywords: Software defect prediction; NASA PROMISE repository; class imbalance; ADASYN; ensemble learning; data leakage; duplicate data; evaluation protocol
Keywords: Software defect prediction; NASA PROMISE repository; class imbalance; ADASYN; ensemble learning; data leakage; duplicate data; evaluation protocol
Abstract
RSAA-UNet: A Residual Squeeze-and-Excitation Attention U-Net with Atrous Context and Focal-Tversky Supervision for Lower-Grade Glioma Segmentation on FLAIR MRI under a Patient-Wise Evaluation Protocol
G. Appasami*, A. Selvakumar
DOI: 10.17148/IJARCCE.2026.151090
Abstract: Outlining lower-grade glioma (LGG) on FLAIR MRI is a prerequisite for volumetric follow-up and radiogenomic work, yet many published U-Net pipelines for the public TCGA-LGG collection do not state, or do not enforce, a clean separation of patients between training and testing. This paper describes RSAA-UNet, a 2-D encoder- decoder that combines residual double-convolution blocks, squeeze-and-excitation (SE) recalibration, additive attention gates on all four skip connections, an atrous spatial pyramid pooling (ASPP) module at the bottleneck, and deep supervision at three decoder levels, trained with a Focal-Tversky plus binary cross-entropy objective. Ten networks (three baselines, two RSAA-UNet variants and five single-component ablations) were trained under one protocol on a patient-wise 78/16/16 split of 110 patients (3,929 slices) and evaluated on 16 unseen patients (498 slices, 182 containing tumour). Flip-based test-time augmentation and small-component removal were examined as inference options. On tumour-containing slices the full RSAA-UNet with both inference steps reached a Dice coefficient of 0.715 (patient-bootstrap 95% CI 0.651–0.771), against 0.714 for U-Net and 0.743 for the strongest baseline, a residual U-Net; no difference was significant at the patient level after Holm correction. The proposed objective lowered the false- positive rate on tumour-free slices to 2.5%, compared with 4.4–6.0% for the baselines, and the ablation pointed to residual connections and SE recalibration as the most useful components, while attention gates and ASPP gave no measurable benefit here. Training of the full model also diverged after epoch 20, so its weights come from an early checkpoint. The Discussion explains why a strict patient-wise protocol gives lower Dice values than those usually quoted for this dataset, compares the results with five published studies, and lists the changes needed before the architecture can be judged conclusively.
Keywords: lower-grade glioma; brain MRI segmentation; U-Net; attention gate; squeeze-and-excitation; atrous spatial pyramid pooling; Focal-Tversky loss; patient-wise evaluation
Keywords: lower-grade glioma; brain MRI segmentation; U-Net; attention gate; squeeze-and-excitation; atrous spatial pyramid pooling; Focal-Tversky loss; patient-wise evaluation
Abstract
AI-Powered Image Captioning and Content Generation: Design, Implementation, and Evaluation
Hirek Praful Warhekar, Kunal Dinesh Afare, Manasvi Dharmendra Ingale, Sagar Sanjay Kinada, Anup G. Kadu
DOI: 10.17148/IJARCCE.2026.151092
Abstract: Image captioning models describe what a photograph shows, but people who publish images online usually need that description reshaped for a particular language, platform, and tone. This paper presents the design, implementation, and evaluation of a web application that bridges this gap. A BLIP captioning model fine-tuned on Flickr8k produces a base caption; a large language model served through the Groq API rewrites it for a selected platform, language (English, Hindi, or Marathi), and writing style; and a Pillow-based renderer places the result on the image. A Streamlit interface, a FastAPI backend, and a SQLite history store complete the system. On 1,000 Flickr8k test images the captioner obtained BLEU 0.2999, METEOR 0.5697, and CIDEr 0.5191. Ten-run timing tests gave a mean of 7.35 s for caption generation and 23.0 s end to end. Limitations of the evidence are discussed.
Keywords: Image captioning; BLIP; vision-language models; large language models; multilingual content generation; Flickr8k
Keywords: Image captioning; BLIP; vision-language models; large language models; multilingual content generation; Flickr8k
Abstract
Design and Validation of a Cognitive 360° PLM Framework for Autonomous and Adaptive Product Lifecycle Decisions
Rahul Chandrayan
DOI: 10.17148/IJARCCE.2026.151093
Abstract: PLM is holistic and cross functional business activity addressing not only product but also organizational structure, working methods, processes, people, information structure and Information system. PLM manages each individual products across its lifecycle this helps companies to have full control over their product and serve the market effectively. The research aims to examine how integrated lifecycle information and adaptive decision-making can support more efficient product management, identify improvement opportunities, and generate measurable business value. A proposed empirical evaluation using relevant industrial case studies, performance data, and comparative analysis could assess the framework's effectiveness against established PLM practices. The expected contribution is a structured approach to connecting PLM integration, data-driven intelligence, continuous adaptation, and value realisation within a single lifecycle management framework. The study seeks to provide both theoretical insights and practical guidance for organisations pursuing operational excellence, sustainable product development, and improved long-term business performance.
Keywords: Product-system; Lifecycle Management; Industry 4.0; Digital Twin, PLM, Business Process Improvement, System Integration, 360° PLM, Adaptive PLM, Data-Driven Decision Support, Lifecycle Optimisation, Return on Investment, Sustainability, Intelligent Analytics.
Keywords: Product-system; Lifecycle Management; Industry 4.0; Digital Twin, PLM, Business Process Improvement, System Integration, 360° PLM, Adaptive PLM, Data-Driven Decision Support, Lifecycle Optimisation, Return on Investment, Sustainability, Intelligent Analytics.
Abstract
Braive: IoT Enabled Smart Braille Learning Glove for Assistive Education
Swathi, Thanmayi Rai A, Thashvi Rai K, Vibha H D, Deepthi M B
DOI: 10.17148/IJARCCE.2026.151094
Abstract: Learning Braille is essential for visually impaired children, yet trained teachers and affordable tactile learning tools are scarce. This paper presents Braive, an IoT-based Braille learning system that pairs a wearable glove with an Android application over Bluetooth Low Energy (BLE). An ESP32 in the glove drives six vibration motors arranged as the dots of a Braille cell and reads six tactile switches for input. The Kotlin application offers a Learning Mode, a five- level gamified Practice Mode with an adaptive question generator, and a word-spelling Test Mode, while speech recognition, text-to-speech feedback and an on-device Room database allow fully offline operation. BLE latency, recognition accuracy and learning outcomes were not measured, and no study with visually impaired learners has yet been carried out. The two-part architecture keeps lesson logic on the phone and real-time actuation on the glove, providing a modular, low-cost approach to Braille education built from commodity components.
Keywords: Braille learning, smart glove, Internet of Things, ESP32, Bluetooth Low Energy, haptic feedback, assistive technology, Android application
Keywords: Braille learning, smart glove, Internet of Things, ESP32, Bluetooth Low Energy, haptic feedback, assistive technology, Android application
Abstract
Artificial Intelligence and its Applications
Sakshi Gorakh Aher, Jitesh Bhimrao Ade, Prof. Gopale. P. J
DOI: 10.17148/IJARCCE.2026.151095
Abstract: Artificial Intelligence (AI) is a branch of computer science concerned with systems that perform tasks associated with human intelligence, including learning, reasoning, perception, language processing, pattern recognition, and decision support. AI is increasingly used in healthcare, education, finance, agriculture, transportation, manufacturing, cybersecurity, and customer service. This paper presents a conceptual, literature-based overview of major AI techniques and their applications. It discusses machine learning, deep learning, natural language processing, computer vision, neural networks, and knowledge-based systems, and outlines a general workflow for selecting, developing, evaluating, and responsibly using AI solutions. The discussion highlights the importance of data quality, appropriate evaluation, privacy, fairness, security, transparency, and human oversight. This is a review paper; no original experimental results are claimed.
Keywords: Artificial Intelligence, Machine Learning, Deep Learning, Natural Language Processing, Computer Vision, Neural Networks, Expert Systems, Responsible AI.
Keywords: Artificial Intelligence, Machine Learning, Deep Learning, Natural Language Processing, Computer Vision, Neural Networks, Expert Systems, Responsible AI.
Abstract
Machine Learning and Explainable AI Techniques for Smart Grid Stability, Security, and Decentralized Energy Management: A Survey
Basima S, Hanna Fathima H, Ninu Bijoy, Sreemayi Saju, Sravana Pradeep
DOI: 10.17148/IJARCCE.2026.151097
Abstract: Smart grids increasingly rely on artificial intelligence to address the variability introduced by renewable energy generation, distributed prosumers, and bidirectional communication in modern power systems. Machine learning (ML), deep learning (DL), and explainable artificial intelligence (XAI) have been applied to various challenges, including rapid stability classification, dynamic security assessment, electricity demand forecasting, and decentralized energy management. These approaches employ diverse algorithms and methodologies, ranging from classical and ensemble learning models to deep neural networks and multi-agent systems. Recent developments also include peer-to-peer (P2P) energy trading, complex-network-based grid resilience analysis, and clustering-based load forecasting. Despite significant progress, several challenges remain, including the heavy reliance on simulated benchmark datasets, limited validation under real-world operating conditions, and the lack of integrated explainability in high-performing predictive models. Differences in datasets, evaluation metrics, and experimental settings further complicate the comparison and generalization of existing approaches. Addressing these limitations requires more diverse datasets, improved model interpretability, stronger cybersecurity measures, and effective integration of grid stability assessment with decentralized energy management and market mechanisms.
Keywords: smart grid stability, machine learning, explainable AI, decentralized energy management, peer-to-peer energy trading, multi-agent systems
Keywords: smart grid stability, machine learning, explainable AI, decentralized energy management, peer-to-peer energy trading, multi-agent systems
Abstract
Low-Cost Wi-Fi CSI-Based Human Activity Recognition Using ESP32 and a Gated Recurrent Unit Classifier
Pavani M B, Priya M S, Vindhya G M, Jeevan H G
DOI: 10.17148/IJARCCE.2026.151098
Abstract: Human Activity Recognition (HAR) has traditionally relied on cameras or wearable sensors, which raise privacy and user-compliance concerns. This paper presents a device-free HAR system based on Wi-Fi Channel State Information (CSI) captured with a single low-cost ESP32 microcontroller, addressing the accessibility gap left by prior work that depends on expensive, multi-antenna, research-grade Wi-Fi hardware. Raw CSI logs were collected from four subjects performing ten activities, parsed, and audited for recording-duration quality; 13 of 150 recordings were excluded as truncated. The remaining data were converted to CSI magnitude, segmented into overlapping one-second sliding windows, and normalized using training-set statistics. A two-layer Gated Recurrent Unit (GRU) classifier with 54,346 parameters was trained with class-weighted cross-entropy loss and early stopping. On a held-out, recording-level test split of 607 windows, the model achieved 64.74% accuracy and 65.38% macro F1-score, and macro F1 rose from 59.19% to 65.38% after the data-quality audit (measured on differently sized test splits). High-motion activities were recognized well (F1 > 0.75), whereas low-motion activities were less separable, which is consistent with the limited resolution of single-antenna CSI. A controlled live test that held the Wi-Fi hotspot constant and changed only the room showed that the learned CSI signatures are room-specific, confirming the cross-environment generalization limitation reported in the literature. The work provides an accessible, reproducible baseline for low-cost CSI-based HAR and an empirical characterization of its present limitations.
Keywords: Wi-Fi Sensing; Channel State Information (CSI); Human Activity Recognition (HAR); ESP32; Gated Recurrent Unit (GRU); Deep Learning; Device-Free Sensing
Keywords: Wi-Fi Sensing; Channel State Information (CSI); Human Activity Recognition (HAR); ESP32; Gated Recurrent Unit (GRU); Deep Learning; Device-Free Sensing
Abstract
CareerCraft AI: An Intelligent Employment Support System Integrating Resume Building, Applicant Tracking System (ATS) Analysis, and GeoBased Job Matching
Samidha S. Shende, Srushti S. Khate, Srushti M. Deshmukh, Anuj G. Gondane, Prof. N. K. Chede
DOI: 10.17148/IJARCCE.2026.151099
Abstract: Job seekers often rely on separate tools for resume analysis, resume enhancement, job-description matching, interview practice, and job search. This paper presents CareerCraft AI, an integrated, candidate-facing web prototype that combines these functions in one workflow. The system uses rule-based extraction for PDF and DOCX resumes and a hosted large language model (LLM), with rule-based fallbacks for selected tasks. The paper describes the architecture, data model, processing workflow, and exploratory observations from a deployed instance. Illustrative outputs demonstrate the available interfaces, while consistency checks identify unresolved differences between displayed scores and simple count-based calculations. Exploratory testing also identified parsing limitations, variation in generated content, and unsupported achievement metrics in some resume rewrites. These observations motivate factual review, transparent score provenance, privacy safeguards, and systematic evaluation. The study documents an integrated prototype; its accuracy, reliability, fairness, and effectiveness have not been established.
Keywords: CareerCraft AI; Applicant Tracking System; Large Language Models; Resume Analysis; Job Matching; Interview Preparation.
Keywords: CareerCraft AI; Applicant Tracking System; Large Language Models; Resume Analysis; Job Matching; Interview Preparation.
Abstract
The Role of Artificial Intelligence in Digital Marketing
Akshada Gorde, Aditya Hilal, Prof Ghodke. G. K
DOI: 10.17148/IJARCCE.2026.1510101
Abstract: Artificial intelligence (AI) is increasingly influencing how organisations plan, deliver and evaluate digital marketing. By analysing large volumes of data, identifying patterns and supporting content generation, AI can help marketers understand audiences, personalise communication, automate routine tasks and evaluate performance. This paper examines AI in digital marketing through a narrative review of selected academic literature, including Bashang and Puttanna (2023) and research on AI-enabled marketing and generative AI. It reviews applications such as customer segmentation, recommendation systems, paid advertising, search and social media optimisation, content creation, conversational agents, predictive analytics and customer relationship management. The paper considers potential benefits, including efficiency, relevance and data-informed decisions, alongside challenges involving privacy, bias, transparency, factual accuracy, cost and brand authenticity. The review suggests that AI is most useful as an augmentation tool rather than a complete replacement for human judgement. Its value depends on reliable data, clear objectives, human review and responsible governance. As a literature-based study, this paper does not claim to have collected primary data or established causal effects; its conclusions are a synthesis of selected sources.
Keywords: artificial intelligence, digital marketing, generative AI, personalisation, marketing automation, customer engagement
Keywords: artificial intelligence, digital marketing, generative AI, personalisation, marketing automation, customer engagement
Abstract
SmartCompress AI: An Entropy-Aware Hybrid Machine Learning Framework for Intelligent Lossless Compression Algorithm Selection
Sanjiv R, Sriramji P, P. Durgadevi
DOI: 10.17148/IJARCCE.2026.1510103
Abstract: Organisations store ever larger volumes of logs, records, source code and backups, yet the choice of compression algorithm for each file is usually left to habit, with one archiver applied to everything. This paper presents SmartCompress AI, a web-based storage optimisation framework that inspects a file before compressing it and recommends the most suitable lossless engine for the user's goal. The system measures Shannon entropy over a streaming 64 KB window, extracts file size, category and extension, and passes these features together with a user preference (fastest, balanced or maximum compression) to a Random Forest classifier. A rule-based engine guards the classifier for pre-compressed and near-random data and acts as a fallback whenever the model is unavailable. Six engines are supported: ZIP, GZIP, BZIP2, LZMA, 7Z and PDF stream re-deflation. On a 17-file, 4.01 MB benchmark corpus, LZMA achieved the highest aggregate saving of 92.28% but needed 567.6 ms in total, about 14 times longer than ZIP (40.8 ms, 90.19% saving). The classifier reached 61.5% held-out accuracy, yet its wrong choices cost only 0.13 percentage points of saving on average, because the candidate engines are close for most compressible data. Every operation is recorded in a SQLite ledger and can be exported as a PDF audit report, and all 34 automated tests pass. The framework demonstrates that a lightweight, explainable recommender can make compression choices transparent and resource-aware, supporting sustainable data storage.
Keywords: Lossless Compression, Shannon Entropy, Algorithm Selection, Random Forest, Hybrid Recommender, LZMA, DEFLATE, Storage Optimisation, Flask, Sustainable Computing
Keywords: Lossless Compression, Shannon Entropy, Algorithm Selection, Random Forest, Hybrid Recommender, LZMA, DEFLATE, Storage Optimisation, Flask, Sustainable Computing
Abstract
Application of Computer Vision Technology in Industrial Automation
Aishwarya Solanke, Bhumika Solanke, Prof. More. N.N
DOI: 10.17148/IJARCCE.2026.1510104
Abstract: Computer vision enables machines to acquire, process, and interpret visual information from industrial scenes, and it has become a key enabling technology for modern automation. This paper presents a structured study of the application of computer vision technology in industrial automation, covering automated visual inspection, defect detection, object detection and localization, vision-guided robotics, quality control, and workplace safety monitoring. A conceptual framework is proposed in which industrial cameras capture images on the production line, preprocessing and feature-extraction stages prepare the data, deep-learning models perform inspection and classification, and the results are passed to programmable logic controllers (PLCs) or robotic controllers for real-time action. Deep-learning approaches such as convolutional neural networks, object detectors, and segmentation networks are discussed together with lightweight edge-deployment models that suit the strict latency requirements of production lines.
The study also examines the main practical challenges, including variation in lighting, limited availability of labelled defect images, real-time processing constraints, integration with legacy control systems, and the need for interpretable decisions. Overall, the framework aims to improve inspection accuracy, consistency, and throughput while reducing manual effort and production waste. However, its actual performance needs to be verified through practical implementation and testing on realistic industrial datasets. Future research should evaluate factors such as detection accuracy, false-alarm rate, inference latency, throughput, and robustness under changing factory conditions to determine its suitability for real-world deployment.
Keywords: Computer vision, industrial automation, deep learning, defect detection, machine vision, Industry 4.0, quality inspection.
The study also examines the main practical challenges, including variation in lighting, limited availability of labelled defect images, real-time processing constraints, integration with legacy control systems, and the need for interpretable decisions. Overall, the framework aims to improve inspection accuracy, consistency, and throughput while reducing manual effort and production waste. However, its actual performance needs to be verified through practical implementation and testing on realistic industrial datasets. Future research should evaluate factors such as detection accuracy, false-alarm rate, inference latency, throughput, and robustness under changing factory conditions to determine its suitability for real-world deployment.
Keywords: Computer vision, industrial automation, deep learning, defect detection, machine vision, Industry 4.0, quality inspection.
Abstract
An AI Powered Automated Code Review System Using Large Language Models and Static Analysis
Mrs. Nita Meshram, C. Balaji Naidu, Ganni Naveen Raj Anudeep, K. Vedavyas, K. Harsha Nandhan
DOI: 10.17148/IJARCCE.2026.1510105
Abstract: Reviewing pull requests by hand is slow and in-consistent, and the automated aids available today each solve only half of the problem: static analysers enforce fixed rules but cannot grasp the context of a change, while large language models (LLMs) can reason about code yet may return inaccu-rate or non-reproducible feedback. This paper surveys work that pairs LLMs with static or program analysis for automated code review, looking at system architectures, the way the two components are combined, the evidence reported, and the de-sign questions that remain open. Ten sources were examined, and five systems with enough accessible detail were set side by side. Combining an LLM with a rule-based or analysis-based component recurs across the set, yet the designs differ in how analysis supports the model, in deployment cost, and in the strength of their evidence. That evidence is uneven: one source gives a precision figure for generated comments, one records favourable reactions from developers, and the rest offer quali-tative or limited detail, so ranking the systems would be unjus-tified. Six open issues are identified: comparable quantitative evaluation, deployment cost and accessibility, merging of con-textual and rule-based findings, verification and reproducibility of model output, handling of failures, and language coverage. A related pull request review system is placed against these is-sues; its software modules are reported as implemented, while end-to-end validation and the final merged report are still in progress. The contribution is therefore a comparison limited to its sources, a gap analysis that stays within the evidence, and a plan for evaluation, not a claim that any system is better.
Keywords: automated code review, large language mod-els, static analysis, Semgrep, pull request, hybrid review
Keywords: automated code review, large language mod-els, static analysis, Semgrep, pull request, hybrid review
Abstract
LLM Red-Teaming and Evaluation Agent System
Chanchal S. Pophale, Mayank A. Darokar, Prashansa P. Patil, Prathmesh S. Tayade, Neha Kandalkar
DOI: 10.17148/IJARCCE.2026.1510106
Abstract: Large Language Models (LLMs) can produce inaccurate, biased, unsafe, or adversarially elicited outputs. This paper presents an LLM red-teaming and evaluation workflow that separates prompt generation, target-model execution, response evaluation, and report generation. The framework covers factual reliability, harmful output, bias and fairness, consistency, prompt injection, and jailbreak resistance. It is designed to apply shared test cases to multiple models and store each prompt, response, evaluation label, and rationale in a structured record. The manuscript reports category-level flag rates for three target models; however, its model-level and category-level tables report different completed-test totals. These results therefore require verification against run-level records. Evaluator flags are treated as screening signals for human review, not as confirmed failures or certification of model safety. The workflow’s evidential value depends on test design, evaluation criteria, and evaluator reliability.
Keywords: Large Language Models, Red-Teaming, AI Safety, Multi-Agent Systems, Prompt Injection, Model Evaluation.
Keywords: Large Language Models, Red-Teaming, AI Safety, Multi-Agent Systems, Prompt Injection, Model Evaluation.
Abstract
Artificial Intelligence and Robotics for Automated Pharmacy and Medicine Dispensing Systems: A Survey
Abhijith T S, D H Aromal, S Nasif, Nithul Das H, Neethu S, Dr.Shani Raj
DOI: 10.17148/IJARCCE.2026.1510107
Abstract: Manual medicine dispensing, drug storage, and pharmacy inventory management remain error-prone, slow, and labor-intensive, motivating a rapid rise in artificial intelligence (AI) and robotics research aimed at automating these processes. This paper presents a structured survey of twenty recent works (2020–2026) spanning robotic drug grasping, vision-based pill/drug identification, automated dispensers and infusion devices, IoT-enabled inventory and supply-chain tracking, and broader reviews of AI/deep-learning techniques underpinning these systems. We propose a four-category taxonomy – Robotic Grasping and Manipulation, Vision-Based Detection and Identification, Automated Dispensing and Delivery Hardware, and IoT/Inventory and Supply-Chain Management – and examine each reviewed paper in terms of technique, dataset, reported results, and trade-offs. A comparative analysis table summarizes accuracy, dataset scale, and system complexity across studies. Our findings show that YOLO-family detectors and hybrid grasp-planning networks dominate vision-guided manipulation, achieving 90–99% detection accuracy, while systematic reviews confirm consistent, though heterogeneous, gains in dispensing efficiency and medication safety from automation. We identify six open challenges, including limited real-world/clinical validation, dataset scarcity, edge-deployment constraints, and lack of standardized benchmarks, and outline future research directions to guide subsequent work in intelligent pharmacy automation.
Keywords: pharmacy automation, robotic grasping, medicine dispensing, YOLO, deep learning, IoT, healthcare robotics
Keywords: pharmacy automation, robotic grasping, medicine dispensing, YOLO, deep learning, IoT, healthcare robotics
Abstract
AI Based Interview and Evaluation System
S Asif Hussain, Mizba, Mohammad Kouser, Mohammed Mehraj Fathima, Vijay Kumar R
DOI: 10.17148/IJARCCE.2026.1510108
Abstract: In today's competitive job market, the interview plays an essential part in determining the candidate's career opportunities. Nevertheless, many candidates experience difficulty preparing for the interview since they do not have enough opportunities for training, are unable to assess their performance adequately, and receive insufficient personalized feedback. Besides, traditional interviewing can take too much time and be biased due to subjective opinion of the interviewer. That is why an innovative solution in the form of VoxEval AI, an AI-based interview evaluation system, which is aimed at assisting candidates in assessing their interview skills and receiving personalized feedback, is suggested. This paper provides information on the development of VoxEval AI – a platform which uses Artificial Intelligence, Machine Learning, Natural Language Processing, Speech Processing, and Computer Vision to analyze various aspects of interview. Candidates will be able to register on the platform, upload their resume, choose a certain position and take mock interviews with the help of artificial intelligence. Resumes will be analyzed for skills and qualifications, answers given by candidates during interview will be evaluated according to their relevancy and quality, and speaking characteristics like clarity, fluency, speaking pace, use of filler words and pauses will be analyzed. On the basis of the assessments, the system will create a performance score, interactive dashboards, and comprehensive reports that enable the candidates to recognize their strong points and weak points. Additionally, the system provides individual suggestions on how to improve the technical knowledge, communication skills, and interview preparation. Through this integration of capabilities in one platform, VoxEval AI strives to make the process of preparing for interviews easier, more structured, and less laborious. With this system, students and job seekers can easily practice their interviews and keep track of their development.
Keywords: VoxEval AI; interview evaluation via artificial intelligence; mock interviews; Artificial Intelligence; Machine Learning; Natural Language Processing; speech recognition; computer vision; resume analysis; automated assessment; personalized feedback.
Keywords: VoxEval AI; interview evaluation via artificial intelligence; mock interviews; Artificial Intelligence; Machine Learning; Natural Language Processing; speech recognition; computer vision; resume analysis; automated assessment; personalized feedback.
Abstract
MediCycle AI: A Smart Framework for Safe and Sustainable Redistribution of Surplus Medicines
Mrs.Anjana H S, Atharva K R, Gnanasagar G N, Jeevan P, Siddeshwar B V
DOI: 10.17148/IJARCCE.2026.1510109
Abstract: Unused and surplus medicines accumulate across households, hospitals, and non-governmental organizations, creating environmental and public-health burdens while access to essential medicines remains uneven. Take-back and reverse-logistics programmes concentrate on collection and disposal governance, demand forecasting targets procurement, and verification, expiry recognition, and AI decision support are demonstrated in isolation. This paper proposes MediCycle AI, an integrated, auditable framework for the safe, expiry-aware, demand-driven redistribution of eligible surplus medicines.It combines image-based verification with OCR for label and expiry extraction; demand forecasting using random forest and XGBoost, with LSTM only where the dataset justifies it; inventory management with near-expiry detection; a transparent multi-factor recipient-matching and priority-scoring model; and controlled generative and agentic AI layers under retrieval grounding and mandatory human approval. Hard safety constraints— expired or unverifiable medicines are never redistributed—are separated explicitly from soft prioritization factors such as demand, urgency, shelf life, quantity, distance, and recipient priority. The paper presents the architecture, database design, protocol, and evaluation metrics. As of submission, implementation and quantitative results are pending; the manuscript reports an evaluation plan and expected outcomes rather than empirical results, and positions the framework strictly as decision support requiring professional and regulatory validation.
Keywords: Surplus Medicine Redistribution, Pharmaceutical Reverse Logistics, Demand Forecasting, XGBoost, OCR, Computer Vision, Medicine Verification, Recipient Matching, Generative AI, Agentic AI, Healthcare Supply Chain, SDG.
Keywords: Surplus Medicine Redistribution, Pharmaceutical Reverse Logistics, Demand Forecasting, XGBoost, OCR, Computer Vision, Medicine Verification, Recipient Matching, Generative AI, Agentic AI, Healthcare Supply Chain, SDG.
Abstract
A Survey on Federated Learning for Privacy-Preserving Healthcare: Towards Differentially Private Early Sepsis Prediction
Harshil P, Sreehari R, Kesiya Shaji, Adivinayak M V, Sukanya M V, Girija V R
DOI: 10.17148/IJARCCE.2026.1510111
Abstract: Hospitals hold huge amounts of sensitive patient data, but privacy rules and data-ownership concerns mean this data usually cannot be pooled to train machine learning models. Federated Learning (FL) gets around this by sending the model to the data instead of the data to the model. Each institution trains on its own records and shares only model updates. In this paper we survey nineteen studies on federated learning in healthcare published between 2022 and 2026, in ascending order of their citation numbers, which follows the order in which they are first mentioned in the Introduction. The studies fall into seven groups: reviews of FL in healthcare, privacy attacks on FL, cross-hospital generalisation of sepsis models, the dataset source, the base paper on federated sepsis early warning, related work on federated sepsis and Intensive Care Unit (ICU) risk prediction, and differential privacy in FL. For every study we note the objective, the method, the main results and the gap it leaves. The picture is mixed. Federated models can beat single-site models in multi-hospital ICU risk prediction, yet only 5.2% of 612 reviewed healthcare FL studies report real clinical use and only 27.7% include an explicit privacy mechanism. Sepsis models can also lose accuracy at new hospitals, and attacks on FL can reveal not only whether a record was used but which client held it. The ten sepsis studies handle privacy very differently from one another: five report no privacy mechanism in the available text, two run clipping or noise tests with no formal guarantee, one tests a single privacy level, one is centralised only, and one never states its privacy budget. As far as we could tell from the available texts, none gives a patient-level privacy guarantee with accounting across federated rounds. We end by listing the open challenges and by placing FedCare, our planned study of differentially private federated early sepsis prediction across two hospital systems in the open PhysioNet/Computing in Cardiology (CinC) 2019 data, against this gap.
Keywords: Federated Learning (FL), Healthcare, Early Sepsis Prediction, Differential Privacy (DP), Differentially Private Stochastic Gradient Descent (DP-SGD), Membership Inference, Non-Independent and Identically Distributed (Non-IID) Data, PhysioNet/Computing in Cardiology (CinC) 2019.
Keywords: Federated Learning (FL), Healthcare, Early Sepsis Prediction, Differential Privacy (DP), Differentially Private Stochastic Gradient Descent (DP-SGD), Membership Inference, Non-Independent and Identically Distributed (Non-IID) Data, PhysioNet/Computing in Cardiology (CinC) 2019.
Abstract
Machine Learning-Based Procurement Price Benchmarking Through Historical Ranking and Market Evidence
Keshava A, Amrutha N V, Anu Patel M V, Deepthi M, Shashank S
DOI: 10.17148/IJARCCE.2026.1510112
Abstract: Historical procurement records can support price benchmarking, but selecting a suitable reference is challenging when similar descriptions represent different brands, models, specifications, commercial units, quantities, or procurement contexts. This paper examines the historical candidate-reranking component developed for procurement price decision support in CivicEngage. For each query, an upstream retrieval stage supplies ten historical candidates. The proposed reranker combines semantic similarity, structured-attribute matching, retrieval signals, and historical price- consensus features using a LightGBM binary classifier with a nested-cross-validation-based decision gate. It either retains the original Rank-1 candidate or replaces it with a candidate from the supplied set; the returned benchmark is the selected candidate's original unit price, not a newly generated or adjusted price. On the documented benchmark of 10,560 queries and 105,600 candidate rows, the gated reranker achieved 38.99% Accuracy@10, compared with 37.57% for the original Rank-1 baseline, an improvement of 1.42 percentage points. On a clean subset of 9,417 queries whose target descriptions did not contain explicit price strings, it achieved 39.70%, compared with 38.36% for the baseline. These results show a measurable but limited improvement for the evaluated candidate pool. They do not establish general current-market price accuracy, and the result remains constrained by whether a suitable historical candidate is retrieved.
Keywords: Procurement Price Benchmarking; Historical Procurement Data; Candidate Reranking; LightGBM; Binary Classification; Semantic Similarity; Price Consensus; Decision Support
Keywords: Procurement Price Benchmarking; Historical Procurement Data; Candidate Reranking; LightGBM; Binary Classification; Semantic Similarity; Price Consensus; Decision Support
Abstract
SentinelGuard: Explainable Detection of Spyware and Remote Access Trojans on Windows
Nirbhay A. Shende*, Vishal V. Ghorad, Soumya P. Mendse, Kanak V. Gupta, Radhika P. Malu
DOI: 10.17148/IJARCCE.2026.1510113
Abstract: Remote access trojans (RATs) and other spyware can secretly use a computer's camera, contact command- and-control (C2) servers and upload data, while signature-based protection may miss new variants. This paper presents SentinelGuard, a software-only monitor for Windows personal computers that learns the normal behaviour of the PC on which it is installed and combines an unsupervised anomaly detector with exact rules. A per-feature normal-range model and an Isolation Forest operate on seven behavioural features, and SHAP values explain every AI alert in plain language; rules cover camera or microphone use by unexpected applications, per-application C2-style connection bursts, unknown local-network devices and strong unknown WiFi networks. To address RAT delivery, a Downloads guard inspects the Windows Mark-of-the-Web of new files and an optional, read-only email monitor checks attachments and links. A self- protecting learner removes rare upload bursts before training and retrains on a rolling 14-day window. On 215 held-out normal readings the false-alarm rate was 1.4%, and five synthetic attack scenarios were all flagged. In live, safe simulations the AI model detected cryptominer-like CPU load (HIGH) and data upload (CRITICAL) with explanations naming the responsible feature, the C2 rule detected every simulated connection burst, a second PC detected all three demonstration steps including camera use, and test RAT-delivery downloads and emails were flagged within 7 s and 13 s. Self-cleaning lowered a learned upload limit inflated by a burst from about 4,201 KB/s to about 298 KB/s. Camera attribution failed on an updated Windows 11 build, and all attacks were simulated.
Keywords: Anomaly detection, explainable artificial intelligence, Isolation Forest, remote access trojan, SHAP, spyware detection, Windows security.
Keywords: Anomaly detection, explainable artificial intelligence, Isolation Forest, remote access trojan, SHAP, spyware detection, Windows security.
Abstract
A Hybrid AI Recommender System for Dynamic Career Guidance
Addhwaith S Ajith, Aaron John Joy, Vaishnav Biju, Ameesha J S, Jency Mariam John, Girija VR
DOI: 10.17148/IJARCCE.2026.1510114
Abstract: The rapid transition from traditional, manual career counseling to dynamic, automated recommendation frameworks represents a vital imperative in modern education and workforce alignment. Conventional career guidance platforms heavily rely on static rule-based psychometric profiling (such as Holland’s RIASEC framework) or isolated machine learning classifiers. However, standalone approaches consistently suffer from critical deficiencies, notably data sparsity, severe cold-start limitations for new candidates, lack of contextual text understanding, and black-box operational opacity. This paper presents a detailed technical architecture and comprehensive research paper on a novel Hybrid AI Recommender System engineered for dynamic career path generation. By unifying supervised classification models (Random Forest, XGBoost, Support Vector Machines), transformer-based Natural Language Processing (BERT embeddings with GRU-driven attention mechanisms), Graph Neural Networks (GNNs) over domain ontologies for Explainable AI (XAI), and a Retrieval-Augmented Generation (RAG) conversational interface, the system establishes a transparent, multi-dimensional decision engine. We formalize the skill gap quantization model using Euclidean distance vector spaces, delineate the technological stack (Python, Scikit-learn, PyTorch, React, Vite, TypeScript), and evaluate the pipeline using standard empirical frameworks. Experimental benchmarks demonstrate that the hybrid approach achieves superior predictive precision (91%), recall (88%), and cold-start error reduction, bridging theoretical psychometrics with dynamic labor market demands.
Keywords: Career guidance, hybrid recommender systems, machine learning, natural language processing, explainable AI, knowledge graphs, RIASEC model, skill gap quantization, semantic matching, RAG architecture.
Keywords: Career guidance, hybrid recommender systems, machine learning, natural language processing, explainable AI, knowledge graphs, RIASEC model, skill gap quantization, semantic matching, RAG architecture.
Abstract
From Fragmented Capability to Unified Investigation: A Critical Review of Artificial Intelligence Across the Police Investigation Lifecycle
Goury S, Anjana V, Nythal A, Roobini R, Girija V R, Rini V R
DOI: 10.17148/IJARCCE.2026.1510115
Abstract: Criminal investigation still relies on manual complaint registration, hours of Closed-Circuit Television (CCTV) review and hand-drafted case files. Artificial intelligence (AI) has been applied to parts of this work, such as crime prediction, criminal network analysis, Automatic Number Plate Recognition (ANPR), face recognition and language-based reporting, but these tools remain disconnected. This paper reviews twenty studies published between 2022 and 2026 and identifies nine recurring gaps, including the absence of end-to-end integration, limited explainability, dataset bias and weak coupling with legal procedure. It then proposes Kaval AI, an officer-in-the-loop investigation assistant that connects seventeen modules through one shared case record, from Malayalam and English voice intake to Bharatiya Nyaya Sanhita (BNS) section prediction and charge sheet drafting, with every suggestion confirmed by an officer. The requirements, architecture, database design and a planned evaluation are presented; as a design-stage proposal, the paper reports no experimental results.
Keywords: Police Investigation, Crime Prediction, Graph Neural Networks, Number Plate Recognition, Face Recognition, Knowledge Graphs, Explainable AI, Bharatiya Nyaya Sanhita, Kaval AI
Keywords: Police Investigation, Crime Prediction, Graph Neural Networks, Number Plate Recognition, Face Recognition, Knowledge Graphs, Explainable AI, Bharatiya Nyaya Sanhita, Kaval AI
Abstract
Evolving Cyber Threats, Cyber Frauds, and Attacks: Strategies for Mitigation
Raju B. Amle
DOI: 10.17148/IJARCCE.2026.1510116
Abstract: Rapid digitalization has increased reliance on interconnected systems, exposing sectors to cyber threats and fraud involving software vulnerabilities, credential theft, social engineering, ransomware, cloud misconfigurations, third- party risks, and AI-generated deception. This narrative review examines emerging threats, their impacts, and a layered defense framework integrating identity security, patching, monitoring, data protection, user awareness, incident response, and recovery. It also considers responsible AI-based threat detection and associated adversarial and privacy risks. Drawing on public guidance and threat reports. This paper presents overview of cybersecurity, cybersecurity threats/attacks cyber frauds, the presents the related work, then describes the Evolution of Cybersecurity, Cyber Threats and Cybercrimes: From 1960 to the Present, Current Cyber Threats and Cyber Frauds, Impact of Cyber Incidents, and along with Strategies for Mitigation.
Keywords: Cyber Threats, Vulnerabilities, Social Engineering, ransomware, cloud misconfigurations, patching.
Keywords: Cyber Threats, Vulnerabilities, Social Engineering, ransomware, cloud misconfigurations, patching.
Abstract
Aegis: An AI-Based Network Intrusion Detection and Prevention System with Stateful Scan Detection and ISO/IEC 27001 Incident Reporting
M. Sreedharan, Lokesh S, Deepashri P, P. Durgadevi
DOI: 10.17148/IJARCCE.2026.1510117
Abstract: Most published machine learning intrusion detectors report accuracy above 99% on CIC-IDS2017, yet much of that figure comes from duplicate flows that end up in both the training and test sets. We built an AI-based network intrusion detection and prevention system (NIDS/NIPS) for Windows hosts that pairs flow classification with a stateful scan detector, decoy honeypot ports and a monitor for failed Windows logons. Alerts are mapped to MITRE ATT&CK techniques and ISO/IEC 27001:2022 controls. Before training, we removed host identifiers and 12,537 exact duplicate flows (22.16% of the data), and fitted the scaler on the training split alone. On the cleaned data XGBoost scored 99.62% accuracy, 99.39% macro F1 and a 0.55% false positive rate, with a median latency of 0.38 ms per flow; Random Forest scored 99.25%, 97.54%, 1.22% and 35.55 ms. When we held out the whole PortScan family, both supervised models called over 89% of those flows benign, and Isolation Forest caught only 0.31%. That result is why the system handles scans with a stateful detector instead of the classifier. Detections feed a four-tier escalation policy that writes Windows Firewall rules, but only after allowlist and subnet checks, with automatic expiry, and in dry-run mode unless an administrator switches enforcement on. A pure-Python flow extractor reproduces 56 CICFlowMeter features exactly and approximates 12 more, so live capture no longer needs Java.
Keywords: Network Intrusion Detection, Intrusion Prevention, XGBoost, Isolation Forest, SHAP, Data Leakage, CIC- IDS2017, Port Scan Detection, MITRE ATT&CK, ISO/IEC 27001
Keywords: Network Intrusion Detection, Intrusion Prevention, XGBoost, Isolation Forest, SHAP, Data Leakage, CIC- IDS2017, Port Scan Detection, MITRE ATT&CK, ISO/IEC 27001
Abstract
A Comparative Study of LWE-Based Post-Quantum Key Encapsulation and Traditional RSA
Rutuja Paraji Zaware, Miss. More N. N.
DOI: 10.17148/IJARCCE.2026.1510118
Abstract: Public-key key establishment protects almost all secure communication on the internet, and RSA remains one of its most widely deployed mechanisms. RSA’s security rests on the difficulty of integer factorisation, a problem that Shor’s quantum algorithm solves in polynomial time on a sufficiently large quantum computer. In response, NIST standardised the lattice-based ML-KEM (FIPS 203), whose security is derived from the Learning With Errors (LWE) problem. This paper presents a comparative study of LWE-based post-quantum key encapsulation and traditional RSA. The study compares the two families on security foundation, quantum resistance, key and ciphertext sizes, and computational cost. To support the comparison with measured data, a Module-LWE key encapsulation mechanism following the ML-KEM-512, ML-KEM-768 and ML-KEM-1024 parameter sets was implemented and benchmarked against RSA-2048, RSA-3072 and RSA-4096 with OAEP padding on the same machine; the study also measures decryption noise margins and end-to-end hybrid encryption. In the experiments, all 900 LWE round trips decapsulated correctly and tampered ciphertexts were rejected. RSA-2048 public keys and ciphertexts were about 2.7 and 3.0 times smaller than those of ML-KEM-512, and RSA encapsulation was far faster; however, RSA key generation was roughly 40 to 300 times slower than LWE key generation, even though the LWE code was an unoptimised Python prototype. The results indicate that LWE-based schemes trade larger messages for quantum resistance and fast key generation, and support a gradual, hybrid migration from RSA.
Keywords: Learning With Errors (LWE), post-quantum cryptography, ML-KEM, RSA, Module-LWE, key encapsulation, quantum resistance
Keywords: Learning With Errors (LWE), post-quantum cryptography, ML-KEM, RSA, Module-LWE, key encapsulation, quantum resistance
Abstract
AI-DRIVEN RESUME ANALYSIS AND PLACEMENT SUPPORT
Anurag AP, Uthara Nair, Sreelekshmi, Amanda Terence, Siji Chacko, Girija VR
DOI: 10.17148/IJARCCE.2026.1510119
Abstract: The rapid adoption of Artificial Intelligence (AI), Machine Learning (ML), and Natural Language Processing (NLP) has expanded the use of automated techniques for resume analysis, recruitment matching, job recommendation, skill development, and interview preparation. This paper presents a comprehensive review of existing research on AI- driven resume analysis and placement support across five interconnected research areas: resume intelligence; semantic matching and Applicant Tracking System (ATS) evaluation; career and skill development; Large Language Model (LLM) and conversational support; and interview intelligence. The reviewed studies employ traditional NLP and machine- learning techniques such as regular expressions (regex), keyword matching, Term Frequency–Inverse Document Frequency (TF-IDF), Bag-of-Words (BoW), Word to Vector (Word2Vec), cosine similarity, k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Logistic Regression and fuzzy logic, together with Named Entity Recognition (NER), Bidirectional Encoder Representations from Transformers (BERT), Robustly Optimized BERT Pretraining Approach (RoBERTa), distilled Bidirectional Encoder Representations from Transformers (DistilBERT), attention-based neural networks, collaborative filtering, retrieval-augmented generation (RAG), and LLM-based architectures. The literature shows a methodological progression from lexical and rule-based processing toward contextual, semantic, and generative approaches, with several studies reporting strong task-specific performance. At the same time, the evidence remains functionally fragmented: resume parsing and matching are extensively studied, while ATS evaluation, skill-gap analysis, learning support, conversational assistance, and AI-based interview preparation are addressed by smaller and more specialized groups of studies. The review identifies an integration gap in the literature, together with limitations related to dataset diversity, evaluation comparability, changing job requirements, explainability, fairness, privacy, and real-world validation. The findings provide a consolidated view of current methods, reported outcomes, and research opportunities in AI-supported resume analysis and placement preparation.
Keywords: Artificial Intelligence, Machine Learning, Natural Language Processing, Resume Parsing, ATS, Semantic Matching, Skill-Gap Analysis, Job Recommendation, Large Language Models, Mock Interview, Placement Assistance.
Keywords: Artificial Intelligence, Machine Learning, Natural Language Processing, Resume Parsing, ATS, Semantic Matching, Skill-Gap Analysis, Job Recommendation, Large Language Models, Mock Interview, Placement Assistance.
Abstract
RansomGuard AI: An AI-Powered Ransomware Detection and Storage Protection System
Rithvik Rajmohan, Roheith R, Bharath R, P. Durgadevi
DOI: 10.17148/IJARCCE.2026.1510120
Abstract: Ransomware continues to pose a severe threat to data availability by executing unauthorized encryption, rapid file alterations, and malicious extension changes across local file storage. This paper introduces RansomGuard AI, an integrated defense-in-depth storage surveillance and recovery system designed to monitor filesystem events, analyze behavioral risk indicators, and maintain cryptographically verified snapshot recovery points. Operating on a Python FastAPI backend paired with a reactive web dashboard, the framework continuously evaluates filesystem activity against key telemetry signals: modification rates, burst write frequencies, Shannon entropy shifts, and file extension mutations. To guarantee disaster recovery without relying on compromised storage, recovery copies are secured using immutable SHA-256 manifest checkpoints. The system includes a sandboxed synthetic attack simulator that reproduces ransomware- like behaviors inside an isolated environment, demonstrating complete monitoring, detection, audit logging, and point- in-time recovery workflows without exposing the host OS to live malware.
Keywords: Ransomware detection, filesystem monitoring, Shannon entropy, behavioral risk profiling, SHA-256 manifests, snapshot recovery, FastAPI, system security.
Keywords: Ransomware detection, filesystem monitoring, Shannon entropy, behavioral risk profiling, SHA-256 manifests, snapshot recovery, FastAPI, system security.
Abstract
From External Grounding to Collective Intelligence: A Comprehensive Survey on Factuality, Knowledge Graphs, and Multi-Agent Orchestration in Large Language Models
Abishek S, Alfred Jaison, Ashish S, Karthik B, Anpu Alexander, Jinu Lenin
DOI: 10.17148/IJARCCE.2026.1510121
Abstract: Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) demonstrate remarkable generative capabilities but suffer from persistent hallucinations, factual drift, and reasoning degradation in high- consequence domains. While canonical Retrieval-Augmented Generation (RAG) mitigates parameter staleness, it introduces critical vulnerabilities including neural ranking miscalibration, numerical insensitivity, and channel noise cascading. This survey provides an exhaustive, taxonomy-driven synthesis of recent advances across four synergistic research pillars: evidential retrieval calibration, information-theoretic entropy bounds, structured knowledge graph (KG) grounding, and multi-agent collaborative orchestration with asymmetric verification against hallucination snowballing. Adhering strictly to PRISMA 2020 guidelines across twenty primary benchmarked architectures and over eighty foundational sources, we formalize the mathematical mechanics of epistemic uncertainty and cascading channel capacity. We present multidimensional comparative matrices evaluating accuracy, computational complexity, and inference latency, systematically analyze open research frontiers, and establish concrete engineering design trade-offs for deploying reliable, graph-grounded multi-agent intelligence in mission-critical environments.
Keywords: Large Language Models, Retrieval-Augmented Generation, Knowledge Graphs, Hallucination Mitigation, Evidential Deep Learning, Multi-Agent Systems, Information Theory
Keywords: Large Language Models, Retrieval-Augmented Generation, Knowledge Graphs, Hallucination Mitigation, Evidential Deep Learning, Multi-Agent Systems, Information Theory
Abstract
AI-Based Smart Classroom Attendance System using Face Recognition with CCTV Camera
Praveen Kallappa Ullagaddi, Rahul K G, Vikhyath, Vishal Paramavva Madar, Prof. Pradeep Kumar K G
DOI: 10.17148/IJARCCE.2026.1510122
Abstract: Manual attendance in classrooms is time-consuming, error-prone and open to proxy marking. This paper presents an AI-based smart attendance system that uses a CCTV camera and face recognition to mark student attendance automatically. Live classroom video is split into frames, faces are detected and converted into numerical face encodings, and the encodings are matched against those of registered students. A recognised student is marked present with date and time, duplicate entries within a session are prevented, and unknown faces are ignored. Records are stored in MongoDB, and a web dashboard supports student registration, live attendance, and Excel report generation. Unit tests of ten core functions all passed. The system offers a contactless, low-cost and scalable alternative to manual roll calls.
Keywords: Face recognition, Smart attendance, CCTV, Computer vision, OpenCV, Artificial intelligence, MongoDB.
Keywords: Face recognition, Smart attendance, CCTV, Computer vision, OpenCV, Artificial intelligence, MongoDB.
Abstract
MediGuard: A Safety-First AI-Assisted Healthcare Companion Platform
Vidhi Ingle, Pratik Wankhede, Sayali Deshmukh, Sujal Jadhav, Prof. R. C. Awatade
DOI: 10.17148/IJARCCE.2026.1510123
Abstract: Access to reliable health information, routine self-monitoring and appropriate specialist guidance remains uneven, and general-purpose chatbots can produce unsafe or overconfident medical statements. This paper presents CareBot AI, a modular web platform that combines a conversational health assistant with a deterministic safety layer, a vitals tracker, a medicine and hydration manager, a medical-report analyser based on optical character recognition (OCR), a rule-based specialist recommender, a hospital finder and a transfer-learning pipeline for X-ray classification. The system uses a Flask REST back end with JWT authentication and SQLAlchemy persistence, and a plain HTML/CSS/JavaScript front end. A central design principle is that emergency detection, measurement-threshold checks and disclaimers are rule- based and independent of the language model, and that every component that is not yet configured reports its status honestly instead of fabricating output. An automated suite of 55 tests, covering authentication, per-user data isolation, input validation, safety behaviour and graceful degradation, passes on the implemented back end. The paper describes the architecture, module design and safety mechanisms, and discusses current limitations and the validation required before any clinical use.
Keywords: Healthcare chatbot, AI health assistant, safety layer, medical report OCR, X-ray analysis, transfer learning, Flask REST API, health monitoring
Keywords: Healthcare chatbot, AI health assistant, safety layer, medical report OCR, X-ray analysis, transfer learning, Flask REST API, health monitoring
Abstract
DESIGN AND PRELIMINARY FUNCTIONAL EVALUATION OF GUARDIAN ATTENDANCE: A BROWSER-BASED FACE RECOGNITION SYSTEM FOR EXAMINATION-HALL ATTENDANCE
Dhanush S, Nirmala Basavaraj Mali, Preethi V M, Rahul N, Nikhil G
DOI: 10.17148/IJARCCE.2026.1510124
Abstract: Attendance recording in examination settings requires a process that is efficient, consistent, and resistant to duplicate entries. Guardian Attendance is a browser-based application developed to support student registration, face-based identity matching, examination workflows, and attendance-record management. The implementation uses a browser facial- analysis library to produce face descriptors and compares descriptors for identity matching; application data persistence is handled through Supabase. This manuscript documents the system design and reports preliminary functional observations from single-user testing. The tests reported by the developer include successful registration, recognition of a registered user, rejection of an unregistered face in one reported trial, and persistence of an attendance record after page reload. A repeated-scan test was also performed. Camera-permission-denial behavior also requires confirmation. No multi- participant dataset or formal accuracy evaluation is currently available. Therefore, the observations are treated as functional checks rather than evidence of general recognition accuracy. The paper defines a reproducible evaluation plan covering known-identity matching, unknown-identity rejection, duplicate attendance, camera-permission handling, processing latency, and privacy safeguards. The current work provides a basis for further controlled testing before any claim of deployment readiness.
Abstract
ROMEO: Raw Onboard Multispectral Earth Observation — An Edge-AI Framework for Latency- and Bandwidth-Efficient Satellite Sensing
Charan M H, Abhishek G, Haridarshan K P, Mallikarjuna, K B Janaki*
DOI: 10.17148/IJARCCE.2026.1510125
Abstract: Wildfire and volcanic-eruption alerts from orbit are delayed less by detection accuracy than by the ground processing chain. This paper reports a working onboard pipeline that analyses raw Sentinel-2 Level-0 granules without any ground-segment product: radiometric calibration, coarse spatial coregistration and coarse georeferencing are applied on board, a spectral gate discards cloudy scenes, thirteen bands are reduced to the three carrying thermal signal, and a 23,714-parameter network quantised to 29.9 KB scans the granule. On an 18-granule catalogue the system reaches 100 % granule accuracy with no false detections, and classifies 240 of 240 unseen patches correctly under leave-granules-out evaluation. Removing coregistration costs 16.67 percentage points, identifying it as the enabling step. Quantisation reduces inference latency eightfold, and transmitting alerts instead of imagery reduces 30.67 MB to 105.74 KB. Because the granules are synthesised, the results characterise pipeline correctness rather than field performance.
Keywords: Coarse spatial coregistration, convolutional neural network, edge computing, integer quantisation, onboard artificial intelligence, raw Level-0 data, Sentinel-2, thermal hotspot detection, wildfire detection.
Keywords: Coarse spatial coregistration, convolutional neural network, edge computing, integer quantisation, onboard artificial intelligence, raw Level-0 data, Sentinel-2, thermal hotspot detection, wildfire detection.
Abstract
AI Powered Sign Language to Text and Speech
Samanvi I, Sanjana Hegde, Shravya K V, Shravya N Shetty, Bharathi K
DOI: 10.17148/IJARCCE.2026.1510126
Abstract: Communication between individuals who use sign language and those unfamiliar with it can be challenging, particularly in everyday interactions. This paper presents an AI-powered sign language recognition system designed to convert hand gestures into text and speech in Kannada, Hindi, and Tamil. The proposed system uses a webcam to capture static hand gestures and MediaPipe to detect hand landmarks and extract spatial features. A Multilayer Perceptron (MLP) model classifies the extracted features into supported sign characters. The recognized characters are then combined to form words and sentences using language-specific text composition rules. Large Language Model (LLM) based text refinement is incorporated to improve the grammatical correctness and readability of the generated text. The refined text is displayed through a user interface and converted into speech using a Text-to-Speech (TTS) module. The proposed system demonstrates the potential of computer vision, machine learning, and natural language processing in supporting multilingual assistive communication and improving accessibility for sign language users.
Keywords: Artificial Intelligence, Sign Language Recognition, Gesture Recognition, Text to Speech, Large Language Model
Keywords: Artificial Intelligence, Sign Language Recognition, Gesture Recognition, Text to Speech, Large Language Model
Abstract
SasyaVedha: A Soil-Aware and Climate-Resilient Agronomic Decision Support System for Indian Farms
Nishanth U, Lalith N, Punith Y D
DOI: 10.17148/IJARCCE.2026.1510127
Abstract: Indian farmers routinely have to make five separate decisions in a single season — what disease is on the leaf, how much yield to expect, when to irrigate, what to feed the soil, and how a coming drought or flood will change all three — and most digital tools answer only one of these at a time. This paper presents SasyaVedha, a web-based agronomic decision support system that brings crop disease identification, yield prediction, irrigation scheduling, fertilizer advisory, and climate-event impact simulation into a single Flask-and-React platform, alongside an enterprise mode for multi-farm operations with role-based access control and an append-only audit log. Disease identification uses a MobileNetV2 transfer-learning CNN trained on the PlantVillage dataset, with a deliberate fallback to a curated disease database when trained weights are unavailable so the feature degrades gracefully rather than failing outright. Yield and irrigation are each modelled with a scikit-learn ensemble — Random Forest for yield, Extra Trees for daily net irrigation need — built inside a Column Transformer pipeline that one-hot encodes crop and soil type and passes four live weather features through unchanged, with weather drawn at request time from the Open-Meteo forecast API. Fertilizer guidance and climate-event projection are implemented as heuristic engines grounded in ICAR-style NPK target ranges and per- state, per-crop vulnerability tables rather than as learned models, a choice made specifically to keep their recommendations traceable to known agronomic thresholds. The system is backed by a file-based SQLite schema covering users, sessions, farms, farm assignments, and an activity log, and the frontend is localized into six Indian languages. We describe the architecture, the methodology behind each of the seven functional modules, the role-based security model, and the engineering trade-offs made along the way, including the current reliance on a bootstrap- synthesized training set for the regression models pending real farm records. We close with the limitations this implies, and the roadmap items identified during development.
Keywords: Agronomic Decision Support; MobileNetV2; Random Forest; Extra Trees; Climate Impact Simulation; Role- Based Access Control; Open-Meteo; Precision Agriculture
Keywords: Agronomic Decision Support; MobileNetV2; Random Forest; Extra Trees; Climate Impact Simulation; Role- Based Access Control; Open-Meteo; Precision Agriculture
Abstract
Agentic Commerce: A Survey of Emerging Landscape
Janhavi Manore*, Sharvari Jamkar, Nupoor Joshi, Sukanya Gupta, Parag Jambhulkar
DOI: 10.17148/IJARCCE.2026.1510128
Abstract: The rapid development of artificial intelligence (AI) has moved electronic commerce from using traditional algorithmic recommendations to relying on autonomous agentic systems. In this new approach, referred to as agentic commerce, large language model (LLM)-powered agents are able to independently discover, assess, negotiate, and carry out financial transactions on behalf of human users. The present survey offers a thorough overview of the developing agentic commerce environment. It looks at the basic architecture and the interoperability protocols that allow smooth machine-to-machine transactions, examines the change in the user experience going from manual browsing to intent- driven delegation, and systematically reviews the existing evaluation frameworks, pointing out the gap between shopping scenarios in simulation and actual real-world performance. The study also investigates the security, privacy, and economic issues specific to this system, such as intent misinterpretation, adversarial manipulation of the marketplace, risks associated with dynamic pricing, and the need for dynamic trust governance. Finally, it identifies the major open challenges and the key research areas that are needed if secure, transparent, and user-aligned agentic commerce platforms are to be developed. This work acts as a fundamental reference for researchers, developers, and industry professionals who are dealing with the next generation of automated digital trade.
Keywords: Agentic commerce, LLM agents, Autonomous transactions, E-commerce protocols, AI security and trust, Agent evaluation frameworks
Keywords: Agentic commerce, LLM agents, Autonomous transactions, E-commerce protocols, AI security and trust, Agent evaluation frameworks
Abstract
A Survey on Privacy-Preserving Face Recognition: Techniques, Challenges and Future Directions
Gopika S Nair, Sebin Sibichan, Sri Sai G Shyam, Stanly Antony V W, Sumi Rehim, Dr. Shani Raj
DOI: 10.17148/IJARCCE.2026.1510129
Abstract: Face recognition (FR) is now widely deployed across authentication, surveillance, and identity management systems. However, because facial data is inherently sensitive, storing, transmitting, and processing it introduces severe privacy and security risks. Cloud-hosted FR architectures, in particular, leave raw facial images and deep feature embeddings vulnerable to data breaches, adversarial manipulation, and unauthorized inference. To address these vulnerabilities, this survey delivers a systematic review of privacy-preserving face recognition (PPFR) techniques published between 2019 and 2026. Analyzing primary studies alongside more than related works, we organize current research into a five-part taxonomy: cryptographic approaches (fully homomorphic, selective, and searchable encryption), distributed and perturbation methods (federated learning and GAN-based anonymization), hardware-level protection (lensless imaging), adversarial robustness (attack and defense methods), and domain-specific applications (blockchain, IoT, and e-voting). For each category, we unpack the core methodology, benchmark datasets, reported accuracy, and the operational tradeoffs among privacy, accuracy, and computational overhead. Our analysis highlights key trade-offs: homomorphic encryption retains near-perfect recognition accuracy but incurs high computational overhead; federated learning and lensless imaging deliver stronger privacy at a moderate cost to accuracy; and adversarial robustness remains a widespread open gap across most PPFR designs. We conclude by outlining key open challenges and future research directions.
Keywords: Privacy-Preserving Face Recognition, Homomorphic Encryption, Federated Learning, Adversarial Robustness, Biometric Template Protection, Lensless Imaging.
Keywords: Privacy-Preserving Face Recognition, Homomorphic Encryption, Federated Learning, Adversarial Robustness, Biometric Template Protection, Lensless Imaging.
Abstract
AutoFix: A Verification-Driven Framework for Automated Program Repair and Security Patching of Python Code
Pranav K, Devabala Kurup S, Rekha P S, Abel Markose, Sidharth S, Soja S Padmam, Girija V R
DOI: 10.17148/IJARCCE.2026.1510130
Abstract: Automated program repair (APR) tools still frequently struggle to generate reliable code. Often, they produce patches that fail to compile, reference undefined names, or trick the system into passing tests simply by deleting core functionality. Furthermore, most of the existing research in security repair is heavily focused on C/C++. In this paper, we review twenty studies covering neural, LLM-based, and vulnerability repair, pinpointing six recurring gaps in the field: uncompilable output, execution-blind training, test-suite overfitting, narrow security coverage, assuming fault locations in advance, and evaluation leakage. To address these limitations, we introduce AutoFix, a verification-driven framework built specifically for Python. AutoFix tracks down and localizes faults by combining AST analysis, tracebacks, and Ochiai ranking. It then maps the underlying weaknesses to their respective CWE categories and uses a LoRA-tuned transformer to generate a ranked list of candidate patches. Before any patch is accepted, it must survive four rigorous sandboxed checks: syntax, compilation, standard tests, and property-based testing. The outcomes of this verification process are then fed back into the system as a graded reward to further fine-tune the model. This paper outlines the framework's overall design, architecture, and evaluation plan, with the actual implementation results set aside for future work.
Keywords: Automated program repair, vulnerability patching, large language models, LoRA fine-tuning, CWE, fault localization, patch verification
Keywords: Automated program repair, vulnerability patching, large language models, LoRA fine-tuning, CWE, fault localization, patch verification
Abstract
Blockchain Based digital forensics for Evidence Collection and Provenance Preservation
Vaishnavi Shankar Kalamkar, Sanika Dattatraya Kalokhe, Miss. Ghodke. G. K
DOI: 10.17148/IJARCCE.2026.1510131
Abstract: Digital evidence is very important in investigating cybercrime and other security-related incidents. However, protecting digital evidence from unauthorized changes and maintaining a clear record of its handling can be difficult. Evidence may pass through several investigators or organizations during an investigation, making it necessary to keep accurate records of how it is collected, transferred, stored, and examined.
This research proposes a blockchain-based framework for collecting digital evidence and maintaining its history securely. The framework uses a permissioned blockchain to record important activities related to digital evidence. Cryptographic hash values help verify whether the evidence has been modified, while blockchain maintains a chronological record of evidence collection, access, transfer, and verification. Only authorized investigators and administrators are permitted to perform specific operations. The original evidence files are stored in a secure external storage system, while the blockchain maintains their related information and integrity records.
The proposed framework aims to make digital forensic investigations more transparent, secure, and accountable. It helps investigators track the movement of evidence and identify unauthorized changes. The design also considers challenges such as data privacy, storage requirements, transaction speed, and the need for proper evidence collection procedures. Overall, this framework provides a foundation for improving digital evidence management and preserving the reliability of evidence throughout the forensic investigation process.
Keywords: Blockchain, Digital Forensics, Digital Evidence, Evidence Preservation, Cryptographic Hash, Chain of Custody.
This research proposes a blockchain-based framework for collecting digital evidence and maintaining its history securely. The framework uses a permissioned blockchain to record important activities related to digital evidence. Cryptographic hash values help verify whether the evidence has been modified, while blockchain maintains a chronological record of evidence collection, access, transfer, and verification. Only authorized investigators and administrators are permitted to perform specific operations. The original evidence files are stored in a secure external storage system, while the blockchain maintains their related information and integrity records.
The proposed framework aims to make digital forensic investigations more transparent, secure, and accountable. It helps investigators track the movement of evidence and identify unauthorized changes. The design also considers challenges such as data privacy, storage requirements, transaction speed, and the need for proper evidence collection procedures. Overall, this framework provides a foundation for improving digital evidence management and preserving the reliability of evidence throughout the forensic investigation process.
Keywords: Blockchain, Digital Forensics, Digital Evidence, Evidence Preservation, Cryptographic Hash, Chain of Custody.
Abstract
Sign-Speak: Real-Time Continuous Indian Sign Language Translator
Rashmi N D, Yashaswini H L, Sinchana M S, Madhushree M, Lavanya N L
DOI: 10.17148/IJARCCE.2026.1510133
Abstract: Communication between users of Indian Sign Language (ISL) and people who do not understand sign language can be challenging. Computer vision and machine learning provide opportunities to develop assistive systems that recognize visual gestures and present their predicted meanings in accessible formats. This paper presents Sign-Speak, a web-based sign-language recognition and translation application that integrates real-time camera input, hand-landmark visualization, gesture classification, text output, speech controls, and uploaded-video processing. The application provides a user interface for viewing the camera feed, observing the predicted sign and confidence score, and accessing translation results. It also supports uploading video files in MP4, AVI, and MOV formats, subject to the interface's displayed 50 MB limit. The system uses a trained machine-learning model within a Python-based application. The implementation combines computer-vision processing and a web interface to provide an accessible way to experiment with sign recognition. The application was launched and its interface and workflows were observed during project testing. Quantitative accuracy, processing speed, and generalization performance must be established through controlled experiments before they can be reported as measured results. The proposed system serves as a prototype for assistive
evaluation.
Keywords: Indian Sign Language, Sign Language Recognition, Computer Vision, Deep Learning, Hand Gesture Recognition, LSTM, Real-Time Processing, Text-to-Speech, Assistive Technology.
evaluation.
Keywords: Indian Sign Language, Sign Language Recognition, Computer Vision, Deep Learning, Hand Gesture Recognition, LSTM, Real-Time Processing, Text-to-Speech, Assistive Technology.
Abstract
An AI-Powered Lie Detection System Using Facial Micro-Expressions
Mrs. Recharla Surendra Geethanjali, Aditya Prakash Masabinal, Bharath Kumar S C, C Yuvaraj, G Daewoo Sri Prasad
DOI: 10.17148/IJARCCE.2026.1510134
Abstract: Deciding whether a person is telling the truth matters in security screening, forensic questioning, recruitment interviews, and fraud inquiries, yet the available instruments are weak. The polygraph demands attached sensors and a trained operator, and unaided human judges perform only marginally above chance. Micro-expressions, which are involuntary facial movements lasting roughly a twenty-fifth to a fifth of a second, give a contactless cue, but they pass too quickly for people to follow reliably. This paper describes the design and prototype of a video-based deception analysis system built around the Real-life Deception Detection Dataset of 121 courtroom clips. It joins facial-landmark geometry with learned visual features, follows how both change over time with a recurrent network that carries an attention layer, and returns a Truth/Lie probability together with an account of the frames and facial regions that shaped it. The design rests on a short review of six recent studies and on informal consultations with a recruiter and a police officer. The upload, analysis, and explained-output interface of the prototype is working, whereas quantitative validation (accuracy, precision, recall, and F1-score) is still to be carried out, so no performance claim is made. The contributions are a documented architecture, an explicit evaluation protocol, and a discussion of ethical limits that treats the tool as support for a human examiner and not as a verdict.
Keywords: lie detection, deception analysis, facial micro-expressions, MediaPipe, CNN-LSTM, attention mechanism, explainable AI
Keywords: lie detection, deception analysis, facial micro-expressions, MediaPipe, CNN-LSTM, attention mechanism, explainable AI
Abstract
AI DRIVEN SUSPICIOUS ACTIVITY RECOGNITION USING DEEP LEARNING
Neelakantappa T T, Meghana G S, Nandini G, Gousiya Fathima A, Anusha R H
DOI: 10.17148/IJARCCE.2026.1510135
Abstract: The growing deployment of closed-circuit television (CCTV) cameras produces large volumes of video that are difficult to monitor continuously through human observation alone. Manual surveillance is vulnerable to fatigue, delayed response, and missed events, while basic motion detection cannot reliably distinguish ordinary movement from potentially suspicious activity. This paper presents a deep-learning-based framework for suspicious human activity recognition from surveillance videos using a hybrid Convolutional Neural Network (CNN) and Long Short- Term Memory (LSTM) architecture. Video frames are extracted and preprocessed using OpenCV. A pretrained CNN, such as MobileNet or InceptionV3, is intended to extract spatial features from individual frames, while an LSTM models temporal relationships across frame sequences. A classification layer predicts activity categories and can provide confidence scores to support alert generation. The framework is designed using Python and TensorFlow/Keras, with potential evaluation on public video datasets such as UCF-Crime or HMDB51. Since measured implementation results were not provided in the project report, this paper does not claim a specific accuracy, frame rate, or real-world performance. Experimental validation is required to establish effectiveness, false-alarm rates, and deployment suitability.
Keywords: Suspicious activity recognition, intelligent surveillance, deep learning, CNN-LSTM, computer vision, human activity recognition, video analytics
Keywords: Suspicious activity recognition, intelligent surveillance, deep learning, CNN-LSTM, computer vision, human activity recognition, video analytics
Abstract
UrbanPulse: A Dynamic Multi-Attribute A* Pathfinding and Real-Time Telemetry Framework for Intelligent Urban Bike Navigation
Arathi H L, Varun P M, S Rahul, Shashank P R, Omkar S
DOI: 10.17148/IJARCCE.2026.1510136
Abstract: Rapid urban motorization and vehicular density have severely escalated safety hazards and physiological friction for non-motorized urban commuters. Commercial navigation systems such as Google Maps and Waze are architected primarily around vehicular throughput, optimizing purely for travel time while routing vulnerable road users (VRUs) onto high-stress arterial corridors, complex grade-separated junctions, and congested roadways. Conversely, existing open-source bicycle routing engines rely predominantly on static topological tagging from OpenStreetMap (OSM), rendering them incapable of adapting to live traffic spillovers, sudden intersection chokepoints, or dynamic road closures. To resolve this paradigm gap, this paper introduces UrbanPulse, a cloud-native, microservices-driven intelligent bike navigation platform. UrbanPulse pioneers a Dynamic Multi-Attribute Cost Model (DMACM) coupled with a Modified A* Pathfinding Engine operating over R-Tree indexed spatial graph representations of urban networks. The engine dynamically ingests real-time vehicular flow telemetry and traffic incidents, synthesizes multi-factor edge weights encompassing segment lengths, velocity degradation, dedicated cycle track segregation bonuses, surface roughness friction, signal latency, and geometric turn penalties derived from vector azimuth differentials. We provide formal mathematical proofs of the heuristic's admissibility and consistency, guaranteeing optimal path convergence. Benchmarked against the dense, mixed-traffic urban road network of Bengaluru, India (|V| ≈ 102,400 nodes, |E| ≈ 248,600 edges), UrbanPulse achieves sub-5ms path computation latency, yields a 38.4% reduction in cyclist exposure to arterial traffic hazards, and demonstrates an 84.2% utilization rate of segregated cycling infrastructure compared to baseline routing systems.
Keywords: Intelligent Transportation Systems (ITS), Bicycle Routing, Modified A* Algorithm, Multi-Attribute Cost Function, OpenStreetMap, Spatial Graph Indexing, Vulnerable Road Users (VRUs), Microservices Architecture.
Keywords: Intelligent Transportation Systems (ITS), Bicycle Routing, Modified A* Algorithm, Multi-Attribute Cost Function, OpenStreetMap, Spatial Graph Indexing, Vulnerable Road Users (VRUs), Microservices Architecture.
Abstract
RESQMESH: AN OFFLINE EMERGENCY COORDINATION SYSTEM
Pushpalatha M, Pavitra N, Ramyashree S S, Nithya S, Sirisha D, Varshini R
DOI: 10.17148/IJARCCE.2026.1510137
Abstract: During disasters and other emergencies, cellular networks and Internet services may become unavailable because of infrastructure damage, congestion, power failure, or limited coverage. This paper presents ResQMesh, an offline-first emergency communication and rescue coordination system that uses nearby Android devices to exchange emergency information through Bluetooth-based communication. Emergency requests contain details such as emergency type, priority, location, battery level, and contact information. Records are stored locally using the Room persistence library and can be forwarded between participating devices. When Internet connectivity becomes available, unsynchronized records are transferred to Firebase Firestore. A web-based administration panel enables authorized administrators to monitor recent, active, and completed cases, inspect case details, and update rescue status. Functional and integration tests documented in the project report covered emergency creation, local storage, device discovery,
test cases passed. ResQMesh is intended to support emergency information transfer and coordination, not to replace professional emergency services or guarantee communication in every environment.
Keywords: ResQMesh, emergency communication, offline-first system, Bluetooth, mesh forwarding, Room Database, Firebase Firestore, rescue coordination.
test cases passed. ResQMesh is intended to support emergency information transfer and coordination, not to replace professional emergency services or guarantee communication in every environment.
Keywords: ResQMesh, emergency communication, offline-first system, Bluetooth, mesh forwarding, Room Database, Firebase Firestore, rescue coordination.
Abstract
A Survey on IoT and Artificial Intelligence-Based Smart Home Energy Management Systems: Techniques, Challenges and Future Directions
Athul Krishna R, Darvesh Muhammed N, Naveen P Jose, Sujith Balachandran, Aswathy S, Dr. Shani Raj
DOI: 10.17148/IJARCCE.2026.1510138
Abstract: The increasing penetration of Internet of Things (IoT) devices, renewable energy resources, smart meters, and connected appliances has created new opportunities for intelligent energy management in residential environments. Conventional energy management approaches are often limited by manual monitoring, inaccurate demand estimation, limited interoperability, and the inability to respond dynamically to changing consumption patterns. This paper presents a technical survey of recent research in IoT- and Artificial Intelligence (AI)-based smart home and smart grid energy management. Twenty representative studies covering energy monitoring, load forecasting, non-intrusive load monitoring, demand response, appliance scheduling, anomaly detection, optimization, edge AI, and user behavior modeling are analyzed. The reviewed studies demonstrate that machine learning, deep learning, optimization algorithms, IoT communication technologies, and edge computing can substantially improve energy awareness, forecasting, cost reduction, and automated control. However, significant challenges remain in dataset availability, cross-household generalization, computational complexity, cybersecurity, privacy, interoperability, and large-scale deployment. The survey develops a taxonomy of existing approaches, compares their methodologies and reported results, identifies research gaps, and discusses future directions toward adaptive, secure, low-cost, explainable, and autonomous home energy management systems.
Keywords: Internet of Things, smart home, energy management, artificial intelligence, load forecasting, demand response
Keywords: Internet of Things, smart home, energy management, artificial intelligence, load forecasting, demand response
Abstract
AI-Powered Automated Code Review for GitHub: A Survey of Static Analysis, Large Language Models, Hybrid Approaches, and Developer Workflows
Archana Raj, Pavithra J, Hima Shaji, Jyolsna K V
DOI: 10.17148/IJARCCE.2026.1510139
Abstract: Automated code review has become an important research direction in software engineering because conventional manual review is time-consuming and dependent on reviewer expertise. Traditional static analysis tools can efficiently identify predefined coding violations and security weaknesses, but they have limited understanding of program intent and may produce false-positive warnings. Recent advances in Large Language Models (LLMs) provide new capabilities for code understanding, defect detection, explanation generation, and automated review. This survey examines recent research on AI-powered automated code review, covering static analysis, machine learning, LLM-based approaches, hybrid analysis, automated program repair, vulnerability detection, and GitHub-based workflows. Twenty research papers are comparatively analyzed with respect to methodology, tools, datasets, evaluation metrics, advantages, limitations, and applicability to automated GitHub code review. The survey identifies important research gaps, including false positives, hallucination, limited contextual understanding, computational cost, language-specific analysis, and limited integration with developer workflows. Based on these findings, the survey motivates a hybrid architecture that combines deterministic static analysis with context-aware LLM reasoning and GitHub Pull Request integration.
Keywords: Automated Code Review, Large Language Models, Static Analysis, GitHub, Software Engineering, Code Intelligence, Code Security
Keywords: Automated Code Review, Large Language Models, Static Analysis, GitHub, Software Engineering, Code Intelligence, Code Security
Abstract
CareBridge – Unified Healthcare Services System
Mr. Anand Rajendra Katariya, Miss. Shreya Gajanan Lawange, Miss. Sania Vivek Raut, Mr. Swaraj Prashant Chandurkar, Miss. Priya Rajesh Bawrani, Dr. Gaurav J. Sawale
DOI: 10.17148/IJARCCE.2026.1510140
Abstract: CareBridge – Unified Healthcare Services System is a web-based application designed to integrate and streamline healthcare services through a centralized digital platform. Traditional healthcare management often involves separate systems for patient registration, appointment scheduling, medical records, laboratory reports, prescriptions, and medicine inventory, which can lead to inefficient coordination and increased administrative workload. To address these challenges, CareBridge provides an integrated solution connecting patients, hospitals, doctors, laboratories, and medical stores. The system enables patients to register, manage their profiles, book appointments, access medical histories, and download prescriptions and laboratory reports. Doctors can manage appointments, maintain patient records, issue prescriptions, and request laboratory investigations. Laboratories can update test statuses and upload diagnostic reports, while medical stores can manage medicine details, monitor stock levels, track batch numbers, and identify expired or near-expiry medicines. The application also supports report generation, PDF downloads, and notifications related to appointments, laboratory reports, prescriptions, and medicine inventory. A modular architecture and structured database facilitate efficient data management and communication between interconnected components. Appropriate authentication and role-based access control can help protect sensitive healthcare information. The proposed system aims to reduce manual paperwork, improve information accessibility, enhance coordination among healthcare stakeholders, and simplify routine administrative activities. CareBridge offers a unified approach to healthcare management and provides opportunities for future enhancement through telemedicine, artificial intelligence, online payments, and advanced healthcare analytics.
Keywords: CareBridge, Unified Healthcare Services System, Healthcare Management, Patient Management, Hospital Management, Appointment Scheduling, Laboratory Management, Prescription Management, Medical Store Inventory, Centralized Database, Web-Based Application.
Keywords: CareBridge, Unified Healthcare Services System, Healthcare Management, Patient Management, Hospital Management, Appointment Scheduling, Laboratory Management, Prescription Management, Medical Store Inventory, Centralized Database, Web-Based Application.
Abstract
Early Detection of Alzheimer’s Disease with Blood Plasma Proteins Using Support Vector Machines
Sharanabasavaraddi B, Spandana KP, Yashaswini LM, Madhuri JK
DOI: 10.17148/IJARCCE.2026.1510142
Abstract: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects memory, cognitive abilities, and daily functioning. Early identification of the disease is important for timely clinical assessment and patient management. This paper presents a machine-learning-based framework for classifying Alzheimer’s disease using biomarker-related data and supervised learning algorithms, with particular emphasis on Support Vector Machine (SVM) and Logistic Regression (LR). The proposed workflow consists of dataset acquisition, data preprocessing, handling missing values, encoding categorical attributes, splitting the dataset into training and testing subsets, model training, and performance evaluation. The models are evaluated using standard classification metrics, including accuracy, precision, recall, sensitivity, and confusion matrix analysis. A web-based interface can make the trained model accessible to users by providing a structured input and prediction workflow. The framework aims to investigate the usefulness of machine learning in supporting the classification of Alzheimer’s disease and to compare the performance of the selected algorithms. However, the reliability of the results depends on the quality, representativeness, and correct identification of the dataset. The proposed system is intended as a research prototype for decision support and not as a replacement for professional clinical diagnosis.
Keywords: Alzheimer’s disease, Blood plasma proteins, Biomarkers, Machine learning, Support Vector Machine, Logistic Regression, Early detection, Classification, Predictive analytics, Healthcare decision support.
Keywords: Alzheimer’s disease, Blood plasma proteins, Biomarkers, Machine learning, Support Vector Machine, Logistic Regression, Early detection, Classification, Predictive analytics, Healthcare decision support.
Abstract
ThreePod: A Verification-First System for Prompt- and Image-Driven 3D Modelling and Printing
Sharvil R. Wange*, Lokesh A. Chaudhary, Jay P. Umbarkar, Hemant N. Humane, Shailesh S. Dhok
DOI: 10.17148/IJARCCE.2026.1510143
Abstract: Generative 3D systems can produce a plausible object from a sentence or a single photograph, but plausibility is not the same as coherence, editability, correct scale, printability, or honest provenance. This paper presents ThreePod, a local-first, verification-first pipeline that treats every generative proposal as an untrusted candidate. The text route plans atomic actions, generates editable Three.js blocks, and commits a block only after schema, grounding, syntax, live- browser, and geometric checks. The image route segments a subject, reconstructs a shape, measures visible-side silhouette agreement, separates observed from inferred surface, and exports a part-labelled GLB. The print route converts an approved result into a simulated, layer-locked toolhead path with material and time budgets. We report an offline suite of 766 tests (723 passed, 2 skipped, 41 deselected, 0 failed), a ten-prompt scene benchmark with 401 geometry checks and zero failures, and four research image assets with valid GLB exports and visible-side IoU between 0.6016 and 0.9640. A CPU-only comparison of 3B and 7B local code models shows the expected quality–latency trade-off. The contribution is not a better-looking generator; it is an auditable pipeline that makes generation inspectable, measurable, and print- aware.
Keywords: 3D generation, text-to-3D, image-to-3D, verification, printability, LLM code generation, provenance, digital manufacturing.
Keywords: 3D generation, text-to-3D, image-to-3D, verification, printability, LLM code generation, provenance, digital manufacturing.
Abstract
AgriCopilot: An Offline-Capable, Edge-Native Single-Page Architecture for Precision Agronomy and Inter-Market Price Routing
Punitha M R, T Ranadeep, V Bhuvan Chandu, Shreyas R V, Yuvan B
DOI: 10.17148/IJARCCE.2026.1510144
Abstract: Smallholder farmers in developing regions face compounding challenges: unreliable soil-crop align-ment, non-transparent advisory systems, post-harvest price collapses caused by seasonal overproduction, and cycle-bound high-interest debt. Existing agri-tech solutions often function in silos, depending heavily on high-cost IoT hardware or continuous cloud connectivity while lacking local language support. This study presents AgriCopilot, an offline-first Single-Page Application (SPA) designed to run edge-native Explainable AI (XAI) directly within client web browsers. Operating through HTML5 Canvas memory buffers, the system executes an image-based soil classification pipeline in under 50 ms without external server calls. The model computes a 9-dimensional color-texture feature vector—comprising CIE- L∗a∗b∗ coordinates, hematite redness metrics, or-ganic carbon indicators, Laplacian roughness, and normalized chromaticities—to categorize soils across six major Indian taxonomic orders. A biophysical stress engine applies non-linear penalties for temperature and precipita-tion anomalies using physiological thresholds established by the Indian Council of Agricultural Research (ICAR), generating a composite Plant Health Score and flagging visual soil mismatches. When critical stress is detected, the engine ranks climate-resilient alternative crops and updates system variables with single-tap interaction. To mitigate localized price drops, a supply-elasticity model detects regional sowing saturation (> 80%), while a geospatial routing engine evaluates road winding factors and freight costs to identify profitable interstate termi-nal markets. Finally, the platform categorizes financial support options (e.g., the 4.0% Kisan Credit Card vs. the Agriculture Infrastructure Fund) based on landholding threshold criteria, delivering explanations through client-side voice synthesis across six regional languages.
Keywords: Explainable Artificial Intelligence (XAI); Precision Agronomy; Edge Computing;Soil Image Classification; APMC Mandi Price Elasticity; Supply-Induced Price Collapse; Inter-Market Price Routing; Targeted Credit Allocation.
Keywords: Explainable Artificial Intelligence (XAI); Precision Agronomy; Edge Computing;Soil Image Classification; APMC Mandi Price Elasticity; Supply-Induced Price Collapse; Inter-Market Price Routing; Targeted Credit Allocation.
Abstract
KubeVision AI: Kubernetes-Native Observability and Root Cause Analysis Platform
Shubham More, Prajwal Mandlecha, Kaustubh Gharat, Kaushal Waykole, Prof. P.A.Jain
DOI: 10.17148/IJARCCE.2026.1510145
Abstract: Managing containerized workloads on Kubernetes simplifies deployment and horizontal scaling, but diagnosing failures in production remains challenging. When a service experiences an issue, errors often propagate across dependent components, triggering waves of correlated alerts that conceal the underlying fault. Conventional Site Reliability Engineering (SRE) workflows depend on bytecode instrumentation agents or sidecar proxies, which introduce runtime latency and memory overhead. At the same time, automated Root Cause Analysis (RCA) techniques based on static thresholds or causal dependency graphs struggle under high-dimensional metric noise and bursty traffic patterns. While Large Language Models (LLMs) offer strong contextual reasoning, single-prompt implementations suffer from hallucinations, context window exhaustion, and alert flooding.
This survey paper reviews current methodologies across cloud-native observability, anomaly detection, and automated root cause analysis. We evaluate telemetry collection mechanisms (bytecode injection, sidecars, and kernel-level eBPF), time-series anomaly detection pipelines, causal dependency discovery, and emerging LLM-based agentic frameworks. We benchmark leading industry platforms (Dynatrace, Datadog, Pixie, K8sGPT) alongside academic systems (CauseInfer, MicroCause, CloudRanger, RCAAgent, TriFleetRCA). To address key research gaps around instrumentation overhead, alert storms, and ungrounded LLM outputs, we propose the architectural design of KubeVision AI: a Kubernetes-native framework combining kernel-level eBPF telemetry, a two-stage anomaly filter (Isolation Forest paired with an LSTM Autoencoder), an alert deduplication gateway, and a multi-agent LangGraph diagnostic pipeline with programmatic reflection gates. Finally, we discuss open research challenges and position time- series capacity forecasting as a planned extension.
Keywords: Kubernetes Observability, Extended Berkeley Packet Filter (eBPF), Root Cause Analysis (RCA), Multi- Agent Systems, Large Language Models (LLMs), Anomaly Detection, AIOps.
This survey paper reviews current methodologies across cloud-native observability, anomaly detection, and automated root cause analysis. We evaluate telemetry collection mechanisms (bytecode injection, sidecars, and kernel-level eBPF), time-series anomaly detection pipelines, causal dependency discovery, and emerging LLM-based agentic frameworks. We benchmark leading industry platforms (Dynatrace, Datadog, Pixie, K8sGPT) alongside academic systems (CauseInfer, MicroCause, CloudRanger, RCAAgent, TriFleetRCA). To address key research gaps around instrumentation overhead, alert storms, and ungrounded LLM outputs, we propose the architectural design of KubeVision AI: a Kubernetes-native framework combining kernel-level eBPF telemetry, a two-stage anomaly filter (Isolation Forest paired with an LSTM Autoencoder), an alert deduplication gateway, and a multi-agent LangGraph diagnostic pipeline with programmatic reflection gates. Finally, we discuss open research challenges and position time- series capacity forecasting as a planned extension.
Keywords: Kubernetes Observability, Extended Berkeley Packet Filter (eBPF), Root Cause Analysis (RCA), Multi- Agent Systems, Large Language Models (LLMs), Anomaly Detection, AIOps.
Abstract
AI-Powered Penetration Testing and Vulnerability Remediation for GitHub Repositories
Mrs. Bindu K.P, Gagana R, Kushal K R, M G Sahana, and M Harshit Pramod
DOI: 10.17148/IJARCCE.2026.1510146
Abstract: AI-Powered Penetration Testing and Vulnerability Remediation for GitHub Repositories presents PentestAI, a web platform that scans a developer's GitHub repositories for security vulnerabilities, verifies the serious findings in an isolated sandbox and proposes fixes as pull requests. The system combines a Large Language Model (Google Gemini) that analyses the most security- relevant source files with a deterministic pattern-matching pass that detects committed secrets such as API keys and private keys.
High and critical findings are re-checked with CodeQL and Semgrep inside network isolated Docker containers to reduce false positives. A fixer service generates a patch with a code model, re-scans the patched code twice and opens a pull request only if the vulnerability is gone and no new issue is introduced. The application is built with TanStack Start, React and Supabase, and supports weekly automated scans, vulnerability lifecycle tracking, reports and an AI security assistant. The results show that the platform turns a manual, expensive security audit into a fast, repeatable and verifiable process for everyday developers.
Keywords: Penetration Testing, Large Language Models, Static Application Security Testing, Vulnerability detection, Secret Detection, Automated Program Repair, Sandboxing
High and critical findings are re-checked with CodeQL and Semgrep inside network isolated Docker containers to reduce false positives. A fixer service generates a patch with a code model, re-scans the patched code twice and opens a pull request only if the vulnerability is gone and no new issue is introduced. The application is built with TanStack Start, React and Supabase, and supports weekly automated scans, vulnerability lifecycle tracking, reports and an AI security assistant. The results show that the platform turns a manual, expensive security audit into a fast, repeatable and verifiable process for everyday developers.
Keywords: Penetration Testing, Large Language Models, Static Application Security Testing, Vulnerability detection, Secret Detection, Automated Program Repair, Sandboxing
Abstract
Measuring Web Tracking and Browser Fingerprinting Across the Popularity Spectrum: A Stratified Crawl of 20,911 Websites from the Tranco Top Million
Krushna Sanjay Mohod, Atharva Sunil Raut, Atharv Rajesh Kakade, Shravani Sachin Yawale, S. R. Gupta
DOI: 10.17148/IJARCCE.2026.1510150
Abstract: Third-party tracking and browser fingerprinting are routinely measured on the most popular websites, yet most of the web lives in the long tail of the popularity distribution. We present a stratified measurement of 20,911 websites drawn from the Tranco top-one-million list in three rank bands (1–10k, 10k–100k and 100k–1M). Each landing page was loaded in an instrumented Chromium browser that routes every connection through a destination-validating safety proxy, records all requests and cookies, matches them against the DuckDuckGo Tracker Data Set, and counts calls to canvas, WebGL and audio fingerprinting APIs. 13,045 pages loaded successfully. Re-weighted to the whole list, 84.1% of websites contact at least one known tracker and 27.0% exhibit at least one fingerprinting signal. Tracker presence is flat across popularity, but invasive practices are not: session replay (17.6% vs 9.6%), fingerprinting (33.3% vs 26.8%) and third-party cookies (55.0% vs 44.2%) are significantly more common at the head of the list than in the tail. We also find that 3,660 sites (17.5%) answered the automated browser with an error or bot-challenge page, and that these challenge pages are themselves heavy finger printers. The resulting feature dataset (107 attributes per site) is used to train the Ovael malicious-tracker detection system.
Keywords: Web tracking, browser fingerprinting, web measurement, stratified sampling, Tranco, privacy, bot detection, dataset.
Keywords: Web tracking, browser fingerprinting, web measurement, stratified sampling, Tranco, privacy, bot detection, dataset.
Abstract
RankSetu AI: AI-Powered Smart Learning, Practice & Exam Readiness Platform
Dr. Suryavanshi Arti P., Vaishnav Gorakh Funde, Shahista Irfan Tamboli, Gaurav Balbhim Jawale, Aditya Raju Jadhav
DOI: 10.17148/IJARCCE.2026.1510151
Abstract: MHT-CET candidates (Maharashtra Common Entrance Test) typically study through video lessons, PDF notes, messaging groups and test-series apps that keep their data in separate silos, which leaves them without level- appropriate practice or dependable feedback. This paper describes RankSetu AI, a mobile-first learning and exam- readiness platform developed with Flutter and Firebase, which links practice, assessment, analysis and improvement into a single cycle. The platform provides subject-wise practice, timed mock tests that are scored automatically, an AI test generator, AI Dost (a Google Gemini-based mentor that solves doubts step by step), weakness detection, test recommendation, two-level caching, single-device sessions, client-side anti-cheat controls and a web-based administration panel. Item Response Theory drives adaptive difficulty, while a normal-distribution model yields a percentile estimate that is always shown with a disclaimer. The paper states openly which modules are implemented, which are planned and what the limitations are.
Keywords: Adaptive Learning, Item Response Theory, AI Mentor, Large Language Model, Mock Test Engine, Weakness Detection, Test Recommendation, Caching, Anti-Cheat Controls, Gamification, MHT-CET, Flutter, Firebase
Keywords: Adaptive Learning, Item Response Theory, AI Mentor, Large Language Model, Mock Test Engine, Weakness Detection, Test Recommendation, Caching, Anti-Cheat Controls, Gamification, MHT-CET, Flutter, Firebase
Abstract
NeoLipi: A Multi-Script Ancient Inscription Recognition and Translation System Using Deep Learning
Ramya Vinayak Naik, Sahana R B, Sakshi Sanjay Hegde, Sathwika M D, Prof. Shrinidhi A
DOI: 10.17148/IJARCCE.2026.1510152
Abstract: Ancient stone inscriptions contain valuable information about the history, culture, language, religion and ad- ministration of earlier periods. However, deciphering such inscriptions are difficult because the characters may be faint, broken, damaged or mixed with the texture of the stone. This paper presents NeoLipi, a multi-script ancient inscription recognition and Kannada translation system developed for Brahmi, Tamil(Vatteluttu) and Devanagari inscriptions. The system uses different recognition methods for the three script groups. The Brahmi pipeline performs image enhancement, binarization, line and character segmentation, stroke reconstruction and CNN-based character recognition. The Tamil pipeline performs binarization, stroke refinement, line seg-mentation, character extraction and CNN-based recognition. The Devanagari pipeline uses grayscale conversion, bilateral denoising, CLAHE, background normalization and an OCR-based recognition module. The recognized text is converted into digital text using Unicode mapping where required and is then translated into Kannada. The system is implemented using Python, OpenCV, NumPy, SciPy, TensorFlow/Keras and Streamlit. The proposed system provides a simple way to process historical inscription images and supports their digital preservation.
Keywords: Ancient inscriptions, Brahmi, Tamil, Devana-gari, CNN, OCR, image processing, Unicode, Kannada translation.
Keywords: Ancient inscriptions, Brahmi, Tamil, Devana-gari, CNN, OCR, image processing, Unicode, Kannada translation.
Abstract
STOCK PRICE PREDICTION
Shree Rakshana R, P. Aarti Pai, Akrithi RJ, Dr. Beenarani Manoj
DOI: 10.17148/IJARCCE.2026.1510153
Abstract: Predicting stock prices is a difficult task due to the volatility and non-linearity of the market, as well as the influence of many factors such as past price trends, trading volume, and investor sentiment. In this project I try to compare machine learning algorithms to see which algorithm predicts the stock prices better from historical data. Historical data (Open, High, Low, Close and Volume) is downloaded from Yahoo Finance (yfinance) and preprocessed by dealing with missing values, scaling and keeping the chronological order. Technical indicators like Moving Average (MA), Relative Strength Index (RSI and MACD) are engineered as additional features. First, three models, Linear Regression, Decision Tree and Random Forest, are trained and tested one after another. These are then combined into an ensemble to see if pooling their predictions improves accuracy. The predictions are compared with the real historical prices and evaluated with RMSE, MAE and R2 score. The predicted prices are translated into BUY/SELL signals to measure practical profitability. The goal of the study is to find the best performing approach, individually or in an ensemble, and to make conclusions to guide investment decisions based on data. A survey of 15 recent literatures shows that tree-based and ensemble methods are strong motivating this comparison.
Keywords: Stock price prediction, Machine learning, Linear Regression, Decision Tree, Random Forest, Ensemble learning
Keywords: Stock price prediction, Machine learning, Linear Regression, Decision Tree, Random Forest, Ensemble learning
Abstract
REVIEW PAPER ON APLICATIONS OF GRAPH NEURAL NETWORKS AND BIOMEDICAL LINK PREDICTION
Atharva Prafulla Pagey, Diya Baldota, Rajlaxmi Take
DOI: 10.17148/IJARCCE.2026.1510154
Abstract: Graph neural networks (GNN) provide a natural framework for learning from relational biomedical data, where drugs, proteins, diseases and adverse events are connected through heterogeneous interactions. This review examines twelve papers spanning foundational Graph Neural Network (GNN) architectures, drug-drug interaction (DDI) prediction models, biomedical knowledge graphs and recent projects in attention, multimodal learning, contrastive learning and cold-start prediction. Architectures like Graph Convoluted Networks (GCN) [1] establish an efficient neighborhood propagation method; Graph Attention Network (GAT) [2] introduces learned neighbor attention; architectures like Relational Graph Convoluted Network (R-GCN) [3] and Relational Graph Attention Networks (r-GAT) [12] incorporate relation-specific transformations. Decagon [4] extends relational graph learning to polypharmacy side effects; and Multimodal Deep Neural Network (MDNN) [5] combines graph and heterogeneous drug features. More recent approaches further incorporate multiple features to current approaches like the Deep Attention Neural Network for Drug- Drug Interations (DANN-DDI) [6], GNN which uses a novel approach of Simplified Molecular Input Line Entry System (SMILES) strings (SmileGNN) [7], or a cross-attention GNN [8], a flow based GNN for emerging drug interations (EmerGNN) [9], a Knowledge Graph based DDI model with contrastive learning (KG-CLDDI) [10], and models which use large-scale biomedical knowledge graphs such as PrimeKG [11]. Rather than comparing numerical scores directly across incompatible datasets, the review focuses on how each architectural choice addresses a specific limitation of earlier approaches. The comparison highlights four recurring requirements for adverse drug event prediction: preserving relation semantics, integrating complementary biological views, handling sparse or unseen entities, and maintaining interpretability and scalability. These observations motivate a three-branch GAT architecture for the proposed ADE project.
Keywords: Drug-Drug Interaction Prediction, Graph Neural Networks, Knowledge Graphs, Link Prediction, Multimodal Learning.
Keywords: Drug-Drug Interaction Prediction, Graph Neural Networks, Knowledge Graphs, Link Prediction, Multimodal Learning.
Abstract
Unmasking Deepfakes: Audio-Visual Deepfake Detection Using AI
Akhila P. V, Sahana J, Sagarika Sujil, and Sreehitha P. C
DOI: 10.17148/IJARCCE.2026.1510155
Abstract: The increasing realism of synthetic and manipulated media has made the authentication of digital video and audio a significant multimedia-forensics problem. Visual-only detectors may miss audio manipulation, while audio-only detectors cannot inspect facial artifacts or temporal inconsistencies. This paper presents a source- grounded description of “Unmasking Deepfakes: Audio-Visual Detection Using AI,” a multimodal detection system documented in the project’s Phase-2 presentation. The visual pathway samples video frames, detects facial regions using YOLOv8, extracts spatial representations using XceptionNet, and models frame-sequence dynamics using a bidirectional long short-term memory (BiLSTM) network. In parallel, the audio pathway extracts the soundtrack, converts it into a Mel-spectrogram representation, and applies a convolutional neural network (CNN) for audio classification. The documented design combines modality-specific information through an attention layer and late-fusion mechanism to produce a Real/Fake prediction and confidence score. A Gradio interface supports video upload and result presentation, while Grad-CAM is identified for visual explanation. The project presentation records the principal processing modules as implemented but states that module/system testing and final performance evaluation remain in progress. Accordingly, this paper contributes a formalized system description, mathematical formulation, inference procedure, and reproducible evaluation plan; it does not claim unverified performance. Dataset identity, split protocol, training configuration, fusion implementation, and measured results must be supplied from experiment logs before quantitative conclusions or deployment claims can be made.
Keywords: Deepfake detection; audiovisual forensics; multimodal learning; YOLOv8; XceptionNet; BiLSTM; Mel- spectrogram; late fusion.
Keywords: Deepfake detection; audiovisual forensics; multimodal learning; YOLOv8; XceptionNet; BiLSTM; Mel- spectrogram; late fusion.
Abstract
AI-BASED SKIN LESION CLASSIFICATION AND REPORTING USING DEEP LEARNING
SharanaBasavaRaddi B, Nirmitha P, Sushmitha H, Chinmai B, Thara R
DOI: 10.17148/IJARCCE.2026.1510156
Abstract: Skin conditions can have similar visual characteristics, making preliminary assessment from images challenging. This paper presents a web-based Skin Disease Detector that combines a trained TensorFlow/Keras image- classification model with a Flask application. A user can create an account, upload a skin image, and receive three ranked class predictions with confidence scores. The system resizes images to 300 × 300 pixels, applies EfficientNet-compatible preprocessing, and passes the processed image to the loaded model. Results are stored in an SQLite database and can be revisited through prediction history. The application also provides disease information and downloadable PDF reports. The work demonstrates an end-to-end software implementation for educational and preliminary screening support. The supplied project files do not include a documented test dataset or evaluation metrics; therefore, no claim of diagnostic accuracy or clinical validation is made. The tool is not a substitute for examination by a qualified healthcare professional.
Keywords: deep learning, EfficientNet, image classification, skin lesions, web application
Keywords: deep learning, EfficientNet, image classification, skin lesions, web application
Abstract
Downfeather: A Machine-Learning-Driven Adaptive Multi-Connection Download Manager
Jatin Verma, Abhishek Jatla, Thanuja N BS, Dr. Beenarani Manoj
DOI: 10.17148/IJARCCE.2026.1510157
Abstract: This paper presents Downfeather, a download manager that keeps learning while a download runs. A file is split into HTTP Range segments and fetched over several connections, and the manager tunes how much each connection reads per call using only the speeds it measures. Development took four iterations. Six recorded traces were clustered with K-Means and mined with Apriori to give a first speed governor; because that governor rested on little data and could only slow a connection down, an online gradient “brain” replaced it. Later versions added exploration noise and finally a saved memory of how speed changed after each of the brain’s own moves. A real download of PixelMaster.exe (251.23 MB) over two connections finished in 43.1 s at 46.64 Mbps, with nearly equal shares (51.5% and 48.5%). A small online controller written with the Python standard library and NumPy can therefore cope with a noisy link without a pre-trained model.
Keywords: Adaptive Download Manager, Multi-Connection Download, Online Learning, Gradient Descent, K-Means, Association Rule Mining, Work Stealing, HTTP Range Requests
Keywords: Adaptive Download Manager, Multi-Connection Download, Online Learning, Gradient Descent, K-Means, Association Rule Mining, Work Stealing, HTTP Range Requests
Abstract
AI-Driven Smart Agriculture Platform for Integrated Crop Health Monitoring and Yield Optimization: A Systematic Literature Review
Rakshith Prabhu, Srihari M.R, Srinivas Gowda C.N, Suhas C, Prof. Anitha L
DOI: 10.17148/IJARCCE.2026.1510158
Abstract: Agriculture is undergoing rapid transformation through the integration of Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and Internet of Things (IoT) technologies. These advancements enable intelligent crop disease detection, crop yield prediction, and precision farming practices that improve agricultural productivity while minimizing resource consumption. Recent studies have demonstrated significant improvements in disease classification accuracy using Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and multimodal learning techniques. Similarly, advanced regression and ensemble learning algorithms have enhanced crop yield prediction by utilizing climatic, soil, and environmental parameters.
This systematic literature review analyzes recent research published between 2021 and 2026 on AI-driven smart agriculture systems. The review compares various machine learning algorithms, datasets, evaluation metrics, advantages, and limitations reported in the literature. Furthermore, the paper identifies current research gaps and discusses future research opportunities involving Explainable Artificial Intelligence (XAI), multimodal learning, federated learning, and real-time agricultural decision support systems. The findings indicate that integrating multiple AI techniques into a unified smart agriculture platform can significantly improve prediction accuracy, transparency, and practical usability for modern precision farming.
Keywords: Artificial Intelligence, Smart Agriculture, Crop Disease Detection, Crop Yield Prediction, Machine Learning, Deep Learning, Explainable AI, Precision Farming.
This systematic literature review analyzes recent research published between 2021 and 2026 on AI-driven smart agriculture systems. The review compares various machine learning algorithms, datasets, evaluation metrics, advantages, and limitations reported in the literature. Furthermore, the paper identifies current research gaps and discusses future research opportunities involving Explainable Artificial Intelligence (XAI), multimodal learning, federated learning, and real-time agricultural decision support systems. The findings indicate that integrating multiple AI techniques into a unified smart agriculture platform can significantly improve prediction accuracy, transparency, and practical usability for modern precision farming.
Keywords: Artificial Intelligence, Smart Agriculture, Crop Disease Detection, Crop Yield Prediction, Machine Learning, Deep Learning, Explainable AI, Precision Farming.
Abstract
Trading Copilot for Cryptocurrency Markets: A Review of Analytics, Prediction, Manipulation Detection, Automated Execution and Conversational Agents, with a Gap Analysis of Existing Systems
Sanskruti Patil, Mokshada Sheth, Shraddha Kokare, Naveli Pangal, Kopal Gangrade
DOI: 10.17148/IJARCCE.2026.1510159
Abstract: Retail cryptocurrency traders operate in a market that is open around the clock, split across many venues, heavy-tailed and exposed to manipulation, yet the tools available to them are divided among charting and alert services, automated bot platforms and research prototypes that rarely work together. This paper reviews the literature and the commercial and open-source landscape relevant to a unified Trading Copilot for crypto markets with the following capabilities: trade analysis, news-based volatility analysis, continuous buy/sell prediction, strategy selection, accumulation–manipulation–distribution (AMD) phase detection, price-distribution and breakdown analysis, cross- exchange price monitoring, automated order placement, notifications and alerts, and a conversational agent. We synthesise the evidence for each capability, compare representative systems, and derive ten gaps. Our main findings are that (i) bot platforms emphasise execution but, in the sources we reviewed, document little analytical or explanatory capability; (ii) LLM trading agents are reported mainly in simulation and without live execution; (iii) manipulation detectors are not coupled to trading decisions; and (iv) conversational LLM agents are reported to retrieve data well but to predict poorly, so they should explain computed outputs rather than forecast. We then describe how a Trading Copilot can close these gaps and propose an evaluation protocol.
Keywords: cryptocurrency, trading copilot, large language models, volatility, market manipulation, Wyckoff method, trading bots, alerts, conversational agent.
Keywords: cryptocurrency, trading copilot, large language models, volatility, market manipulation, Wyckoff method, trading bots, alerts, conversational agent.