A CNN-RNN System for Real Time Credit Card Fraud Detection Using Three-Way Classification to Reduce False Positives /Negatives and Detect Fraud Across All Transaction Values

Financial institutions and consumers whose accounts are compromised suffer financial loss from credit card fraud caused by traditional fraud detection systems which rely on static, rule-based mechanisms which are limited in adaptability to evolving fraud techniques, generation of high false positives, false negative and Inadequate capacity to detect low-value fraud-transactions under £50 that are increasingly exploited for “card testing.” These limitations point out a critical need for a system that can accurately detect fraudulent transactions of any value in real time, minimise false positives and negatives, and adapt to emerging patterns. This research aims to design a real time CNN-RNN hybrid fraud detection system with three way classification to reduce false positives and negatives and detect low and high value credit card fraud.