International Journal of Advanced Research in Computer and Communication Engineering

Abstract: This paper presents an integrated credit card fraud classification approach based on ensemble learning and deep learning techniques to address the challenges of class imbalance and evolving fraud patterns in financial transaction data. Ensemble models such as Random Forest and Gradient Boosting are employed to enhance prediction reliability by combining multiple base classifiers, while deep learning architectures, including Deep Neural Networks, are utilized to capture complex and non-linear relationships within transaction features. Comprehensive data preprocessing and imbalance handling strategies are applied to improve model robustness. The proposed framework is evaluated using real-world credit card transaction datasets and assessed through standard performance metrics. Experimental results demonstrate that the hybrid ensemble–deep learning model outperforms traditional machine learning classifiers in terms of accuracy, precision, recall, F1-score, and area under the ROC curve. The findings confirm that the proposed approach provides an effective, scalable, and reliable solution for real-time credit card fraud detection in modern digital payment systems.
The rapid growth of digital payment systems has significantly increased the volume of credit card transactions, making fraud detection a critical challenge for financial institutions. Credit card fraud classification aims to accurately distinguish fraudulent transactions from legitimate ones while minimizing false alarms. This task is particularly complex due to the highly imbalanced nature of transaction data, evolving fraud patterns, and the need for real-time decision-making. In this work, machine learning-based classification techniques are employed to analyze transaction behavior and identify potential fraud. Preprocessing steps such as data normalization, feature selection, and imbalance handling are applied to improve model performance. Multiple classifiers are trained and evaluated using standard performance metrics including precision, recall, F1-score, and area under the ROC curve. The experimental results demonstrate that intelligent classification models can effectively detect fraudulent activities with high accuracy and reduced false positives. The proposed approach enhances transaction security and supports financial organizations in mitigating monetary losses while ensuring a seamless experience for genuine customers.

Downloads: | DOI: 10.17148/IJARCCE.2026.151134