Enhancing Credit Card Fraud Detection in Banking Using Neural Networks

Abstract

Credit card fraud is one of the most serious threats in today’s digital banking systems, where thousands of transactions occur every second. Detecting fraudulent transactions in real time is challenging due to the high imbalance between legitimate and fraudulent records and the constantly evolving nature of fraud patterns. In this project, a deep learning-based fraud detection system is developed using the TabNet model, which efficiently handles large-scale, high-dimensional tabular transaction data. The model learns to focus on the most important features of each transaction using sequential attention, achieving high precision and recall while minimizing false alarms. The system integrates advanced pre-processing, balanced data handling, model training, evaluation, and visualization. Finally, a Flask-based web application is built to provide an interactive user interface with modules for registration, login, real-time prediction, and visual analytics of model performance. The proposed system achieves excellent accuracy, faster inference, and robust fraud detection capabilities compared to traditional approaches.