INTELLIGENT CREDIT CARD FRAUD PREVENTION USING GRAPH NEURAL NETWORKS AND AUTOENCODER MODELS

Abstract

Credit card fraud has become a major concern in modern digital banking due to the rapid growth of online transactions and electronic payment systems. Traditional fraud detection methods often rely on rule-based systems or conventional machine learning algorithms, which may struggle to detect complex and evolving fraud patterns in real time. To address these challenges, this study proposes an intelligent credit card fraud prevention framework using Graph Neural Networks (GNN) and Autoencoder models. The proposed approach leverages the ability of Graph Neural Networks to capture relationships and interactions between entities such as users, transactions, and merchants. By representing transaction data as a graph structure, the model can identify suspicious patterns and hidden connections that may indicate fraudulent activities. Additionally, an autoencoder-based anomaly detection mechanism is employed to learn normal transaction behavior and detect unusual deviations that may signal potential fraud. The integration of graph-based learning and deep anomaly detection enhances the system’s capability to detect both known and unknown fraud patterns. The proposed framework processes transaction data in near real time, enabling faster and more accurate fraud identification while reducing false positives. Experimental evaluation demonstrates that the combined use of GNN and autoencoder models significantly improves fraud detection performance compared to traditional methods. This intelligent fraud detection system can assist financial institutions in strengthening security, minimizing financial losses, and improving trust in digital banking services by providing an effective solution for real-time credit card fraud prevention.