← Back to Archives
VOLUME 15, ISSUE 6, JUNE 2026
Publication in progress..
Multi-Agent AI Systems for Security Automation
Mohammed Kashif, Abdul Rahman Jibran Syed
Integration of SIEM Data Analytics and AI for Proactive Cyber Threat Hunting
Abdul Hasham, Ramesh Venkata Sai lakshmi
Design Evaluation and Validation of a Resilient IoT-Based Flood Prediction Framework for Data-Scarce Environments in East Africa
Muwanga Erasto Kosea, Dr Otanga Daniel, Dr. Satwinder Singh Rupra
QUANTIFYING EXPLANATION DRIFT UNDER MODEL COMPRESSION IN CLINICAL RISK PREDICTION
Stow May Tamara, Maudlyn Ireju Victor-Ikoh
Design and Analysis of a Low-Power High- Speed 32-bit ALU Using Optimized CMOS Architecture
Dr. Kishore M, Dr. Dileep J, Dr. K. Senthil Babu
Directing the Algorithmic Edge of Inclusive Pedagogy: A Comprehensive Review of AI- Driven Assistive Technologies in Education
Dr. Anju Kaushik, Dr. Anil Kaushik
Smart Security System in Train Using Arduino, Ultrasonic Sensor and Camera
HumeraBano Asadullah Baig
Agentic AI for Data Analysis: A RAG-Enhanced Local LLM Framework with Adaptive Visualization
Harsh Mahesh Tatmute, Shivraj Sunil Shinde, Karan Adinath Nemane, Sandesh Sunil Pujari, Rahul Sudhir Ranjane, V. G. Khetade
CivicAi Complaint Management System
Sammed Mangave, Vaibhav Rajput, Ayan Mujawar, Ganesh Hipparkar, Prof. Dr. V.V.Kheradkar
AI-Driven Malicious URL Detection Using Graph Neural Networks
Shraddha Pailwan, Avishkar Patil, Pranav Patil, Yash Shinde, Gayatree Jadhav, Prof. A. B. Majgave
“An Explainable Hybrid LSTM–Random Forest Framework for Accurate Pulmonary Disease Detection and Classification”
Ajay Pal Singh, Ankita Nigam
CodEzy: An AI-Powered Competitive Coding and Personalized Learning Platform
Kunal Sunil Magdum, Akshay Sachin Mulay, Subahan Riyaj Mulla, Shreyansh Nitin Patil, Prasad Sunil Sutar, Prof. (Dr.) V.V.Kheradkar*
Cyberbullying Detection in Social Media Contents using Machine Learning Techniques
Amey Gujar, Akhilesh Ghorpade, Indrajeet Chougule, Vedant Gawas, Paras Gurjar, Himanshu Baboria, Prof. Vinod Khetade
Seasonal Crop Price Prediction using LSTM
Amol Sawant, Dr. Pravin Patil
Evaluating Usability Techniques in Modern Web Applications
Simran, Dr Pooja Rana
Blockchain Based Certificate Verification System
Shivaprasad M S, Usha M
Impact Of Deep Learning Techniques on Super Resolutions
Deepali Karajgikar, Abhishek Magar
A Graph Retrieval-Augmented Generation Framework for AI-Powered Supplier Discovery
Dr.B.Vijayalakshmi, Dr.M.Swarna Sudha, Dr.K.Vijayalakshmi
Abstract
Multi-Agent AI Systems for Security Automation
Mohammed Kashif, Abdul Rahman Jibran Syed
DOI: 10.17148/IJARCCE.2026.15601
Abstract: Considering the increasing popularity of MAAI solutions for providing a promising alternative to automating the process of security operations in complex network environments. The paper presents a scalable multi- agent solution that involves autonomous intelligent agents working together to discover threats, recognize security gaps, take decisions, and respond to security attacks. Unlike other solutions, which involve conventional and centralized security frameworks [9–13], our multi-agent-based approach makes use of several agents, providing more flexibility due to the possibility of running tasks in parallel, adapting quickly by learning agents from experience, and making the whole system less vulnerable to advanced cyberattacks. Our multi-agent system uses agents with unique features and works using the communication layer, in which agents exchange information about detected threats and optimize their actions to achieve the best results in always-on protection. The experimental prototype is tested for its efficiency in detecting DDoS attack, phishing or malware intrusions. The research results revealed high levels of efficiency demonstrated through high detection rates, low response latency, and fewer false positives. As far as MAAI systems go, this work proves a considerable potential for automation and scalability.
Keywords: Multi-Agent Systems; Cybersecurity Automation; Artificial Intelligence; Intrusion Detection System (IDS); Distributed Security; Autonomous Agents; Threat Detection; Security Orchestration Network Security Intelligent Systems.
Keywords: Multi-Agent Systems; Cybersecurity Automation; Artificial Intelligence; Intrusion Detection System (IDS); Distributed Security; Autonomous Agents; Threat Detection; Security Orchestration Network Security Intelligent Systems.
Abstract
Integration of SIEM Data Analytics and AI for Proactive Cyber Threat Hunting
Abdul Hasham, Ramesh Venkata Sai lakshmi
DOI: 10.17148/IJARCCE.2026.15602
Abstract: The ever-shifting landscape of cyber threats is always on the move, with APTs, insider attacks, and zero- day attacks at the forefront. Conventional, rule-based SIEMs—until now, the workhorse of many security operations— demonstrate their limitations in the face of such threats. They can wade through massive amounts of security data, but they tend to vomit out lots of false positives and lack the ability to predict what’s around the corner. This research investigates how SIEMs might do more than simply respond to threats: it examines the use of AI and other forms of advanced analytics to predict future intrusions. To improve the signal-to-noise ratio, add context, and accelerate response times, the proposed solution relies on behavioral analytics, anomaly detection, machine learning, and automated threat intelligence enrichment. The research describes an analytics workflow, an AI-based SIEM solution, and a methodology for comparing these AI-infused systems to traditional systems. The findings indicate that as AI continues to evolve, AI-based SIEM solutions enable organizations to concentrate on threats that matter, minimize the need for continuous human interaction, and identify complex or unexpected attacks earlier.
Keywords: Intrusion Detection Systems (IDS), anomaly detection, alert prioritization, cyber threat intelligence, security information and event management (SIEM), predictive threat hunting, behavioral analytics, data analytics, and machine learning.
Keywords: Intrusion Detection Systems (IDS), anomaly detection, alert prioritization, cyber threat intelligence, security information and event management (SIEM), predictive threat hunting, behavioral analytics, data analytics, and machine learning.
Abstract
Design Evaluation and Validation of a Resilient IoT-Based Flood Prediction Framework for Data-Scarce Environments in East Africa
Muwanga Erasto Kosea, Dr Otanga Daniel, Dr. Satwinder Singh Rupra
DOI: 10.17148/IJARCCE.2026.15603
Abstract: Many Internet of Things (IoT)-based flood prediction systems deployed in developing regions fail to deliver reliable early warnings due to unreliable sensors, fragmented datasets, and limited operational resilience. While numerous frameworks have been proposed, few studies systematically evaluate their design limitations or validate enhanced solutions under realistic failure conditions. This paper presents the design evaluation, enhancement, and validation of a resilient IoT-based flood prediction framework. Using Design Science Research principles, existing IoT flood prediction frameworks were evaluated using ITIL-aligned governance criteria to identify deficiencies in data reliability, service continuity, and system governance. An enhanced framework was then designed and validated through simulation using CHIRPS rainfall data and controlled sensor failure scenarios. Simulation results indicate that the enhanced framework maintains prediction accuracy between 82.4% and 91.6% under increasing data-loss conditions and improves alert timeliness compared to baseline approaches. The findings indicate that resilience-oriented, data-centric IoT design significantly improves flood prediction performance in resource-constrained environments.
Keywords: Flood prediction; Internet of Things; design science research; data reliability; sensor unreliability; early warning systems.
Keywords: Flood prediction; Internet of Things; design science research; data reliability; sensor unreliability; early warning systems.
Abstract
QUANTIFYING EXPLANATION DRIFT UNDER MODEL COMPRESSION IN CLINICAL RISK PREDICTION
Stow May Tamara, Maudlyn Ireju Victor-Ikoh
DOI: 10.17148/IJARCCE.2026.15604
Abstract: Predictive models on tabular clinical data increasingly use SHAP and LIME explanations, while compression is routine. This study quantifies how compression affects post hoc explanations on five clinical benchmarks. A multilayer perceptron was trained, then subjected to L1 pruning at four sparsities (30 to 90 percent) and quantization to four bit widths (8 to 2 bits), yielding nine variants. SHAP, LIME, and permutation importance were applied to each variant and compared to the full model. The transparency cost of compression is compression-type-dependent: heavy pruning generally degrades both accuracy and explanations together, so accuracy alone catches the problem; heavy quantization more often preserves accuracy while degrading explanations, so accuracy alone misses the problem. On two of five datasets at 2 bit quantization, AUC retention exceeds 0.92 while SHAP rank correlation against the full model falls below 0.70. Explanation fidelity should be reported alongside accuracy specifically when quantization is used.
Keywords: Explainable Artificial Intelligence, SHAP, LIME, Model Compression, Clinical Risk Prediction
Keywords: Explainable Artificial Intelligence, SHAP, LIME, Model Compression, Clinical Risk Prediction
Abstract
Design and Analysis of a Low-Power High- Speed 32-bit ALU Using Optimized CMOS Architecture
Dr. Kishore M, Dr. Dileep J, Dr. K. Senthil Babu
DOI: 10.17148/IJARCCE.2026.15605
Abstract: With continuous scaling of CMOS technology and increasing demand for energy-efficient computing systems, the design of high-performance and low-power arithmetic circuits has become a critical challenge in Very Large-Scale Integration (VLSI). The Arithmetic Logic Unit (ALU) is one of the most frequently utilized components in microprocessors, digital signal processors, and embedded systems. Its power consumption and delay significantly influence overall system performance. This paper presents the design, implementation, and analysis of a 32-bit low- power high-speed ALU using an optimized Carry Lookahead Adder (CLA) architecture in 45nm CMOS technology. The proposed design incorporates transistor sizing optimization, clock gating, and multi-threshold CMOS techniques to minimize dynamic and leakage power. Performance evaluation is carried out in terms of propagation delay, average power consumption, and Power-Delay Product (PDP). Comparative analysis with a conventional Ripple Carry Adder (RCA)-based ALU demonstrates significant improvements in speed and energy efficiency. The proposed architecture is suitable for next-generation low-power computing applications.
Keywords: VLSI, CMOS, Low-Power Design, ALU, Carry Lookahead Adder, Power-Delay Product, 45nm Technology, Clock Gating.
Keywords: VLSI, CMOS, Low-Power Design, ALU, Carry Lookahead Adder, Power-Delay Product, 45nm Technology, Clock Gating.
Abstract
Directing the Algorithmic Edge of Inclusive Pedagogy: A Comprehensive Review of AI- Driven Assistive Technologies in Education
Dr. Anju Kaushik, Dr. Anil Kaushik
DOI: 10.17148/IJARCCE.2026.15606
Abstract: This research paper explores the transformative integration of Artificial Intelligence (AI) and human-computer interaction within inclusive digital education, focusing extensively on higher education institutions (HEIs). By synthesizing contemporary literature, structured inventories of EdTech tools and applied case studies, we analyse how AI-powered screen readers, voice assistants, speech recognition software and Natural Language Processing (NLP) interfaces pull apart traditional learning barriers for students with visual, physical and cognitive disabilities.
The findings demonstrate that AI fundamentally shifts assistive frameworks from rigid, linear and high-dependency human-mediated systems to dynamic, contextual and highly autonomous multi-modal learning environments. However, this pedagogical revolution introduces complex infrastructure requirements, acute data privacy issues, algorithmic drop- off and socio-cultural vulnerabilities—including Generative AI Addiction Syndrome (GAID) and technostress.
Keywords: Artificial Intelligence, Higher Education, Natural Language Processing, EdTech
The findings demonstrate that AI fundamentally shifts assistive frameworks from rigid, linear and high-dependency human-mediated systems to dynamic, contextual and highly autonomous multi-modal learning environments. However, this pedagogical revolution introduces complex infrastructure requirements, acute data privacy issues, algorithmic drop- off and socio-cultural vulnerabilities—including Generative AI Addiction Syndrome (GAID) and technostress.
Keywords: Artificial Intelligence, Higher Education, Natural Language Processing, EdTech
Abstract
Smart Security System in Train Using Arduino, Ultrasonic Sensor and Camera
HumeraBano Asadullah Baig
DOI: 10.17148/IJARCCE.2026.15607
Abstract: This research paper presents an innovative approach to railway safety through the development of an Arduino Nano-based train accident prevention system. The system utilizes ultrasonic sensors for real-time obstacle detection on railway tracks, Camera Module for identifying the object, coupled with immediate alert mechanisms including audible buzzers and visual LED indicators. A distinctive feature of this implementation is the integration with Processing IDE software, which provides a graphical interface displaying real-time obstacle detection data for train operators. The prototype demonstrates effectiveness within a 1-meter detection range, offering a cost-effective solution (â‚ą3500 approx) compared to conventional railway safety systems. The paper comprehensively covers the system design, implementation challenges, test results, and proposes future enhancements including IoT integration and machine learning applications for improved reliability under various environmental conditions.
Keywords: Railway Safety, object detection, Camera Module, Arduino Nano, Ultrasonic Sensor, Real-time Monitoring, Obstacle Detection, Embedded Systems
Keywords: Railway Safety, object detection, Camera Module, Arduino Nano, Ultrasonic Sensor, Real-time Monitoring, Obstacle Detection, Embedded Systems
Abstract
Agentic AI for Data Analysis: A RAG-Enhanced Local LLM Framework with Adaptive Visualization
Harsh Mahesh Tatmute, Shivraj Sunil Shinde, Karan Adinath Nemane, Sandesh Sunil Pujari, Rahul Sudhir Ranjane, V. G. Khetade
DOI: 10.17148/IJARCCE.2026.15608
Abstract: This paper presents an Agentic AI Data Analyst model that uses the RAG technique in combination with local large language models deployed using Ollama for privacy protection and low costs associated with intelligent data analysis. This system is no longer dependent on cloud APIs but is capable of providing high quality analysis due to improved incorporation of domain knowledge into the process. The model is coordinated via multi-step reasoning techniques facilitated by LangChain and supported by FAISS vector storage systems for efficient domain knowledge searching. A FastAPI application server runs in the background, providing connectivity to a React front end interface that allows dual modes of data visualization.
Regarding the performance analysis, five open-source language models were tested – LLaMA 3.2 (3B), LLaMA 3.1 (8B), Mistral 7B, Gemma 3 4B, and Gemma 3 12B. In addition, the assessment was conducted in terms of five core criteria: quality of responses, accuracy of the charts generated, depth of insights, rate of hallucinations, and time required for inference. Each of these models was evaluated with and without RAG implementation to determine the exact effects of retrieval augmentation. According to the results of the experiments, RAG offers a performance increase from 25% to 39% in addition to reducing the number of hallucinations by 62-75%. Most notably, RLaMA with 3B parameters and RAG always outperforms Gemma with 12B parameters but without RAG. These results confirm the hypothesis that retrieval of structured domain knowledge is more effective than scaling the number of model parameters. This research shows that lightweight agentic systems locally installed on a computer could serve as a viable alternative to commercial AI solutions.
Regarding the performance analysis, five open-source language models were tested – LLaMA 3.2 (3B), LLaMA 3.1 (8B), Mistral 7B, Gemma 3 4B, and Gemma 3 12B. In addition, the assessment was conducted in terms of five core criteria: quality of responses, accuracy of the charts generated, depth of insights, rate of hallucinations, and time required for inference. Each of these models was evaluated with and without RAG implementation to determine the exact effects of retrieval augmentation. According to the results of the experiments, RAG offers a performance increase from 25% to 39% in addition to reducing the number of hallucinations by 62-75%. Most notably, RLaMA with 3B parameters and RAG always outperforms Gemma with 12B parameters but without RAG. These results confirm the hypothesis that retrieval of structured domain knowledge is more effective than scaling the number of model parameters. This research shows that lightweight agentic systems locally installed on a computer could serve as a viable alternative to commercial AI solutions.
Abstract
CivicAi Complaint Management System
Sammed Mangave, Vaibhav Rajput, Ayan Mujawar, Ganesh Hipparkar, Prof. Dr. V.V.Kheradkar
DOI: 10.17148/IJARCCE.2026.15609
Abstract: Rapid urbanization has led to an increase in civic infrastructure issues such as potholes, garbage accumulation, waterlogging, damaged streetlights, and drainage problems. Traditional complaint management systems often face challenges such as delayed reporting, manual categorization, inefficient complaint tracking, and lack of transparency in the resolution process. This paper proposes CivicAI, an AI-Powered Civic Complaint Management System designed to improve the efficiency of reporting, classifying, tracking, and resolving civic complaints. The system enables citizens to submit complaints by uploading images along with location information through a web-based platform. Artificial Intelligence and Computer Vision techniques are utilized to analyze complaint images, identify the type of civic issue, and determine its severity level. The generated complaint reports are automatically forwarded to the appropriate municipal authorities for further action. The system also provides real-time complaint tracking, status notifications, resolution updates, and analytical dashboards for effective complaint management. By integrating Artificial Intelligence, Computer Vision, Geolocation Services, and Web Technologies, CivicAI enhances transparency, improves response time, and strengthens communication between citizens and municipal authorities, thereby contributing to efficient urban governance and smart city development.
Keywords: Artificial Intelligence, Computer Vision, Civic Complaint Management, Smart City, Geolocation Services, Urban Governance, Image Classification.
Keywords: Artificial Intelligence, Computer Vision, Civic Complaint Management, Smart City, Geolocation Services, Urban Governance, Image Classification.
Abstract
AI-Driven Malicious URL Detection Using Graph Neural Networks
Shraddha Pailwan, Avishkar Patil, Pranav Patil, Yash Shinde, Gayatree Jadhav, Prof. A. B. Majgave
DOI: 10.17148/IJARCCE.2026.15610
Abstract: The exponential proliferation of internet-based services has been accompanied by a parallel surge in cyber threats, particularly the distribution of malicious Uniform Resource Locators (URLs). Phishing attacks, financial fraud, and data exfiltration perpetrated through deceptive URLs cause billions of dollars in losses annually. Conventional detection mechanisms—predicated on static blacklists and rule-based filters—exhibit inherent limitations in identifying novel, obfuscated, or zero-day threats. This paper presents an AI-driven URL classification framework that harnesses the representational power of Graph Neural Networks (GNNs) to model inter-URL relational dependencies alongside individual lexical and structural URL features. In the proposed architecture, URLs are encoded as graph nodes and semantic or behavioral relationships between them are captured as weighted edges. A Graph Convolutional Network (GCN) is subsequently trained on a composite dataset aggregated from PhishTank, ISCX-URL2016, and Kaggle malicious URL repositories. Experimental evaluation on a balanced 80/20 train-test split yields an accuracy of 96.8%, precision of 97.1%, recall of 96.4%, and F1-score of 96.7%, outperforming baseline Support Vector Machine (SVM), Random Forest (RF), and Multi-Layer Perceptron (MLP) classifiers by margins of 4–9 percentage points. The system exposes a Flask-based REST API and lightweight web interface, enabling real-time single-URL and batch classification. Results corroborate the hypothesis that relational graph-based modelling substantially improves detection efficacy and generalization, with particular gains on obfuscated and previously unseen URL patterns.
Keywords: URL detection; graph neural network; graph convolutional network; phishing detection; cybersecurity; malicious URL classification; deep learning.
Keywords: URL detection; graph neural network; graph convolutional network; phishing detection; cybersecurity; malicious URL classification; deep learning.
Abstract
“An Explainable Hybrid LSTM–Random Forest Framework for Accurate Pulmonary Disease Detection and Classification”
Ajay Pal Singh, Ankita Nigam
DOI: 10.17148/IJARCCE.2026.15611
Abstract: Pulmonary diseases such as Chronic Obstructive Pulmonary Disease (COPD), pneumonia, and lung cancer continue to be leading causes of global morbidity and mortality. Timely detection and accurate diagnosis are essential for effective treatment and improved clinical outcomes. Traditional diagnostic techniques—relying heavily on chest X- rays and CT scans—are often constrained by manual interpretation, which is time-consuming and susceptible to human error. This paper proposes a novel hybrid diagnostic framework integrating Long Short-Term Memory (LSTM) networks with Random Forest (RF) ensemble learning to improve the detection and classification of pulmonary conditions. LSTM networks are employed to capture temporal dependencies in sequential clinical data, while the RF model enhances classification robustness and accuracy. The proposed approach includes comprehensive preprocessing of medical imaging and structured clinical data, feature extraction, and model training on an extensive annotated dataset. Evaluation metrics such as accuracy, sensitivity, specificity, and F1-score reveal that the LSTM-RF hybrid outperforms conventional machine learning models. Furthermore, Explainable AI (XAI) techniques are incorporated to ensure model interpretability, promoting transparency in clinical decision-making. The study also highlights real-world deployment challenges, including data privacy, algorithmic bias, and regulatory compliance. The key contributions of this research lie in the integration of deep learning with ensemble techniques and the emphasis on explainability, making it a viable solution for real-time pulmonary disease diagnosis in clinical settings.
Keywords: Artificial Intelligence, Machine Learning, Deep Learning, Pulmonary Disease Detection, Hybrid LSTM, Random Forest, Explainable AI, XAI, COPD, CT scan.
Keywords: Artificial Intelligence, Machine Learning, Deep Learning, Pulmonary Disease Detection, Hybrid LSTM, Random Forest, Explainable AI, XAI, COPD, CT scan.
Abstract
CodEzy: An AI-Powered Competitive Coding and Personalized Learning Platform
Kunal Sunil Magdum, Akshay Sachin Mulay, Subahan Riyaj Mulla, Shreyansh Nitin Patil, Prasad Sunil Sutar, Prof. (Dr.) V.V.Kheradkar*
DOI: 10.17148/IJARCCE.2026.15612
Abstract: Programming education has evolved significantly with the emergence of online learning platforms and competitive coding environments. However, most existing systems continue to rely on binary evaluation techniques that assess only the correctness of program outputs. Such approaches fail to provide meaningful insights into code quality, algorithmic efficiency, readability, and adherence to software engineering principles. This limitation often prevents learners from understanding their mistakes and improving their programming skills effectively. This paper presents CodEzy, an AI-powered competitive coding and personalized learning platform designed to enhance programming education through adaptive learning, intelligent code evaluation, gamified engagement, and real-time coding competitions. The proposed system integrates personalized tutorials, coding challenges, AI-generated feedback, performance analytics, and skill-based coding duels within a unified ecosystem. Unlike conventional platforms, CodEzy evaluates code beyond correctness by analyzing efficiency, structure, style, and maintainability. The platform employs secure Docker-based sandbox execution, cloud-ready architecture, Redis-powered caching, and external Large Language Models (LLMs) for semantic code analysis and educational content generation. Experimental analysis indicates that the platform can provide detailed AI feedback within acceptable response times while maintaining low- latency interactions for competitive coding environments. The proposed system supports educational institutions, coding bootcamps, and self-learners through adaptive learning, intelligent feedback, and integrated coding practice.
Keywords: Artificial Intelligence, Adaptive Learning, Competitive Programming, Code Evaluation, Gamification, Educational Technology, Learning Analytics, Large Language Models (LLM).
Keywords: Artificial Intelligence, Adaptive Learning, Competitive Programming, Code Evaluation, Gamification, Educational Technology, Learning Analytics, Large Language Models (LLM).
Abstract
Cyberbullying Detection in Social Media Contents using Machine Learning Techniques
Amey Gujar, Akhilesh Ghorpade, Indrajeet Chougule, Vedant Gawas, Paras Gurjar, Himanshu Baboria, Prof. Vinod Khetade
DOI: 10.17148/IJARCCE.2026.15613
Abstract: Cyberbullying is a serious problem in the Information Age. It spoils people's sentiments and wellbeing with ugly messages and cruel words. There's so much content on social media at all times, that it would be hard to find this stuff manually as it would take you a long time and you can't expand that easily. Therefore, the researchers tried to come up with a great solution - a Machine Learning framework that automatically detects cyberbullying. It employs NLP methods to clean up the text, such as normalizing words, tokenizing text and interpreting emojis. Plus, it can handle English, Hindi, Marathi and Hinglish texts as well!
Once the text is sorted, the system converts this information to numbers, known as TF-IDF. Then, it employs a Linear Support Vector Machine for classification, using sklearn’s svm.SVC(linear) kernel. There were several different SVM setups that were considered during development, but the linear SVM proved to have the greatest accuracy and computational requirement.
Our experiments demonstrate that the TF-IDF and Linear SVM model is quite effective in the classification tasks with a lesser amount of resources and is efficient. We ran it on a sample of 31,183 text messages from social media, with 23,820 of them classified as bullying and 7,363 as safe. The one thing that makes our system stand out is its multiple language processing and ability to recognize emojis. This allows it to handle the numerous modes of communication on social media. Moreover, we used it as a Flask based API, so it can be integrated with Web apps easily. Ergo, it is a convenient instrument for in real life content moderation and to improve the safety online.
Keywords: Cyberbullying Detection, Machine Learning (ML), Natural Language Processing (NLP), Text Classification, Support Vector Machine (SVM), TF-IDF, Sentiment Analysis, Multilingual Text Processing, Social Media Analysis, Flask API, Emoji Processing, Online Safety.
Once the text is sorted, the system converts this information to numbers, known as TF-IDF. Then, it employs a Linear Support Vector Machine for classification, using sklearn’s svm.SVC(linear) kernel. There were several different SVM setups that were considered during development, but the linear SVM proved to have the greatest accuracy and computational requirement.
Our experiments demonstrate that the TF-IDF and Linear SVM model is quite effective in the classification tasks with a lesser amount of resources and is efficient. We ran it on a sample of 31,183 text messages from social media, with 23,820 of them classified as bullying and 7,363 as safe. The one thing that makes our system stand out is its multiple language processing and ability to recognize emojis. This allows it to handle the numerous modes of communication on social media. Moreover, we used it as a Flask based API, so it can be integrated with Web apps easily. Ergo, it is a convenient instrument for in real life content moderation and to improve the safety online.
Keywords: Cyberbullying Detection, Machine Learning (ML), Natural Language Processing (NLP), Text Classification, Support Vector Machine (SVM), TF-IDF, Sentiment Analysis, Multilingual Text Processing, Social Media Analysis, Flask API, Emoji Processing, Online Safety.
Abstract
Seasonal Crop Price Prediction using LSTM
Amol Sawant, Dr. Pravin Patil
DOI: 10.17148/IJARCCE.2026.15614
Abstract: Predicting agricultural crop prices is a critical task that helps farmers and agricultural stakeholders make informed decisions regarding cultivation, storage, and marketing strategies. Crop prices are influenced by various factors, including seasonal changes, weather conditions, market demand, and supply fluctuations. Due to the complex and dynamic nature of these factors, accurate forecasting remains a significant challenge. This project proposes a Seasonal Crop Price Prediction system using Machine Learning and Long Short-Term Memory (LSTM) networks.
The system analyzes historical crop price data along with environmental and market-related parameters such as temperature, rainfall, humidity, production levels, and previous price trends. LSTM, a powerful recurrent neural network architecture, is capable of capturing temporal patterns and long-term dependencies within time-series data, enabling more precise price predictions. The developed model assists farmers, traders, and policymakers by providing early insights into future market prices, helping reduce risks and improve planning. By leveraging advanced deep learning techniques, the proposed solution contributes to the development of data-driven and sustainable agricultural practices
Keywords: Seasonal Crop Price Prediction, LSTM, Machine Learning, Deep Learning, Time-Series Forecasting, Smart Agriculture, Agricultural Data Analysis.
The system analyzes historical crop price data along with environmental and market-related parameters such as temperature, rainfall, humidity, production levels, and previous price trends. LSTM, a powerful recurrent neural network architecture, is capable of capturing temporal patterns and long-term dependencies within time-series data, enabling more precise price predictions. The developed model assists farmers, traders, and policymakers by providing early insights into future market prices, helping reduce risks and improve planning. By leveraging advanced deep learning techniques, the proposed solution contributes to the development of data-driven and sustainable agricultural practices
Keywords: Seasonal Crop Price Prediction, LSTM, Machine Learning, Deep Learning, Time-Series Forecasting, Smart Agriculture, Agricultural Data Analysis.
Abstract
Evaluating Usability Techniques in Modern Web Applications
Simran, Dr Pooja Rana
DOI: 10.17148/IJARCCE.2026.15615
Abstract: Globally, the use of web-based applications is expanding quickly. Users' requirements for accessibility and engagement have consequently evolved dramatically. Conventional webpages are no longer sufficient. Users have a new experience with modern web applications. Nowadays, developers and the industry place a high value on usability. Every day, more people utilize web apps, and they favor user-friendly and comprehensible platforms. Nonetheless, many programs remain to have usability problems. This paper also examines the advantages and disadvantages of usability evaluation techniques. Currently, IT is becoming an integral part of daily life, with people seeking fast and simple solutions. In response to user needs, various usability techniques have been created. Several factors influence user experience with these techniques. This study aims to examine user behaviour and usability methods using a straightforward survey cross-sectional method, including a sample of users interacting with web applications.
Keywords: usability, web applications, user experience, evaluation, testing.
Keywords: usability, web applications, user experience, evaluation, testing.
Abstract
Blockchain Based Certificate Verification System
Shivaprasad M S, Usha M
DOI: 10.17148/IJARCCE.2026.15616
Abstract: Validating academic and professional credentials efficiently remains a critical security and administrative challenge for global institutions. Traditional verification methods rely heavily on manual verification workflows or centralized databases that lack real-time public access, scale poorly, and are vulnerable to singular points of failure, unauthorized tampering, and permanent data loss. The absence of a unified, low-latency, and tamper-proof verification ecosystem leaves corporate and educational sectors exposed to credential fraud and escalating administrative evaluation overhead.
To address these vulnerabilities, this paper introduces the proposed system, an open-source, decentralized platform that revolutionizes credential management by mapping certificates to unique Non-Fungible Tokens (NFTs) on the high- throughput Sui blockchain while storing physical document assets across the distributed Walrus storage protocol. This project implements an asynchronous decoupled processing pipeline where structural metadata is managed through Move smart contracts, and cryptographic file identifiers (Blob IDs) are stored over decentralized storage arrays. This architecture enables permissionless, zero-account public verification with sub-second latency, alongside transparent, on- chain revocation mechanisms that ensure a permanent audit trail. Empirical testing demonstrates optimal transaction efficiency, highly scalable storage performance via dynamic epoch handling, and absolute resistance to linguistic or historical tampering.
Keywords: Sui Blockchain, Move Smart Contracts, Non-Fungible Tokens, Walrus Protocol, Decentralized Storage, Cryptographic Verification.
To address these vulnerabilities, this paper introduces the proposed system, an open-source, decentralized platform that revolutionizes credential management by mapping certificates to unique Non-Fungible Tokens (NFTs) on the high- throughput Sui blockchain while storing physical document assets across the distributed Walrus storage protocol. This project implements an asynchronous decoupled processing pipeline where structural metadata is managed through Move smart contracts, and cryptographic file identifiers (Blob IDs) are stored over decentralized storage arrays. This architecture enables permissionless, zero-account public verification with sub-second latency, alongside transparent, on- chain revocation mechanisms that ensure a permanent audit trail. Empirical testing demonstrates optimal transaction efficiency, highly scalable storage performance via dynamic epoch handling, and absolute resistance to linguistic or historical tampering.
Keywords: Sui Blockchain, Move Smart Contracts, Non-Fungible Tokens, Walrus Protocol, Decentralized Storage, Cryptographic Verification.
Abstract
Impact Of Deep Learning Techniques on Super Resolutions
Deepali Karajgikar, Abhishek Magar
DOI: 10.17148/IJARCCE.2026.15617
Abstract: Image Super-Resolution (SR) is an important research area in image processing that focuses on reconstructing high-resolution (HR) images from low-resolution (LR) images. The objective of super-resolution is to recover lost details, improve image quality, and generate visually enhanced images. Traditional interpolation methods such as nearest- neighbor, bilinear, and bicubic interpolation often fail to preserve fine details, edges, and textures, resulting in blurred outputs.
With the advancement of Deep Learning, especially Convolutional Neural Networks (CNNs), significant improvements have been achieved in image reconstruction tasks. This research presents a study on the impact of deep learning techniques in image super-resolution, focusing on CNN-based architectures including Super-Resolution Convolutional Neural Network (SRCNN), Fast Super-Resolution CNN (FSRCNN), Very Deep Super-Resolution Network (VDSR), and Enhanced Deep Residual Network (EDSR).
The study analyzes the working principles, advantages, and limitations of these models. Experimental implementation demonstrates that deep learning-based methods can effectively learn complex mappings between low-resolution and high-resolution images, producing sharper edges, improved textures, and better visual quality. However, advanced architectures require higher computational resources and larger datasets for training.
Keywords: Image Super Resolution, Deep Learning, Convolutional Neural Network, SRCNN, FSRCNN, VDSR, EDSR, Image Processing
With the advancement of Deep Learning, especially Convolutional Neural Networks (CNNs), significant improvements have been achieved in image reconstruction tasks. This research presents a study on the impact of deep learning techniques in image super-resolution, focusing on CNN-based architectures including Super-Resolution Convolutional Neural Network (SRCNN), Fast Super-Resolution CNN (FSRCNN), Very Deep Super-Resolution Network (VDSR), and Enhanced Deep Residual Network (EDSR).
The study analyzes the working principles, advantages, and limitations of these models. Experimental implementation demonstrates that deep learning-based methods can effectively learn complex mappings between low-resolution and high-resolution images, producing sharper edges, improved textures, and better visual quality. However, advanced architectures require higher computational resources and larger datasets for training.
Keywords: Image Super Resolution, Deep Learning, Convolutional Neural Network, SRCNN, FSRCNN, VDSR, EDSR, Image Processing
Abstract
A Graph Retrieval-Augmented Generation Framework for AI-Powered Supplier Discovery
Dr.B.Vijayalakshmi, Dr.M.Swarna Sudha, Dr.K.Vijayalakshmi
DOI: 10.17148/IJARCCE.2026.15618
Abstract: As modern supply chains demand higher resilience, agility, and visibility, the task of supplier discovery becomes increasingly critical. We present a novel AI-powered methodology that combines Graph Neural Networks (GNNs), Retrieval-Augmented Generation (RAG), and Large Language Models (LLMs) to enhance supplier search and reasoning. A structured Supplier Capability Knowledge Graph (SCKG) is built by extracting domain-specific triplets from unstructured manufacturing data using fine-tuned LLMs and is enriched through semantic normalization via ontology and manufacturing thesaurus. A GNN-based retrieval system identifies relevant subgraphs by performing dense reasoning over the SCKG. These subgraphs are verbalized into natural language using shortest-path reasoning chains and fed into an LLM for generative explanation. To improve retrieval precision, a hybrid entity normalization technique leveraging Jaccard similarity and vector-based retrieval is applied. This integrated GNN-RAG system significantly outperforms traditional and zero-shot LLM-based supplier search approaches in both precision and recall on real-world datasets. Our results demonstrate the system's ability to perform robust, real-time supplier discovery while enabling explainable and accurate responses.
Keywords: Supplier Discovery, Knowledge Graphs, GNN-RAG, Large Language Models, Semantic NormalizationI
Keywords: Supplier Discovery, Knowledge Graphs, GNN-RAG, Large Language Models, Semantic NormalizationI