A credit card fraud detection system based on Credit-Attention-GRU network

Credit card fraud is a significant global issue, involving unauthorized use of payment cards to obtain goods or services, driven by the increasing digitization of financial transactions. In response to the increasing prevalence of credit card fraud, we propose a novel fraud detection system based on a Multi-Head Attention (MHA) mechanism combined with a Gated Recurrent Unit (GRU) network. The proposed system leverages the strengths of MHA to effectively capture the complex temporal dependencies and patterns in credit card transaction like transaction frequency, amount fluctuations, and changes in geographic location. These patterns reflect normal user behavior, and any significant deviation may indicate potential fraud, while GRUs enable the system to maintain long-term memory and mitigate the vanishing gradient problem. Through extensive experimentation on real-world datasets, we demonstrate that our Credit-Attention-GRU model outperforms state-of-the-art methods in terms of precision, recall, and F1-score. This study contributes to the field of fraud detection by providing a robust and accurate solution for identifying fraudulent credit card transactions.