The Roadmap to Digital Transformation in Banking: Advancing Credit Card Fraud Detection with Hybrid Deep Learning Model

Credit card fraud is a rising concern in the financial sector, driven by the rapid expansion of digital payment systems and the sophistication of fraudulent schemes. Ensuring the security of transactions is critical for protecting consumers, maintaining financial system integrity, and fostering trust in digital payment systems. Existing fraud detection techniques often struggle with the highly imbalanced nature of transaction datasets and evolving fraud tactics, leading to suboptimal detection accuracy and high false positive rates. To address these challenges, this study proposed a hybrid deep learning model that combines an attention mechanism with a multi-layer Gated Recurrent Unit network. The GRU architecture effectively captures sequential dependencies in transaction data, while the attention mechanism improves the model's emphasis on important features indicative of fraud. The research utilized a dataset comprising 284,807 European credit card transactions and Synthetic Minority Oversampling Technique is used to balance the dataset. The hybrid DL model attained 98.76% accuracy, 98.01% precision, and 97.81% recall, ensuing in an F1-score of 97.90%. The ROC-AUC score of 0.983 further demonstrated its capability to efficiently differentiate between fraudulent and legitimate transactions. These results underscore the model's ability to address dataset imbalance and adapt to evolving fraud patterns, offering a reliable solution for real-time fraud detection. This approach not only minimizes financial losses but also strengthens consumer confidence in digital transactions, supporting the continued growth of the global financial ecosystem.