Credit Card Fraud Detection

Credit card fraud has become a major concern in modern digital banking systems due to the rapid growth of online financial transactions. Financial institutions process a large number of transactions every second, making it difficult to identify fraudulent activities accurately and in real time. One of the key challenges in fraud detection is the highly imbalanced nature of transaction datasets, where fraudulent transactions represent only a small fraction of the total data. In addition, fraud patterns continuously evolve, making traditional detection techniques less effective. To address these challenges, this research proposes a deep learning–based fraud detection framework using the TabNet architecture, which is specifically designed for tabular data. The model utilizes a sequential attention mechanism to automatically select the most relevant transaction features, improving both prediction accuracy and interpretability. The system incorporates data preprocessing, class balancing techniques, model training, evaluation, and visualization to enhance detection performance. Furthermore, the trained model is integrated into a Flask-based web application that provides an interactive interface for user authentication, real-time fraud prediction, and performance analysis. Experimental results demonstrate that the proposed system achieves high accuracy, improved precision and recall, and reliable fraud detection performance compared to conventional approaches.