REAL-TIME CREDIT CARD FRAUD DETECTION USING INTEGRATED GNN AND AUTOENCODER MODELS

Abstract

Autoencoders and Graph Neural Networks (GNN) are two advanced deep learning methods that will be employed in this study to enhance fraud detection in financial systems and prevent real-time credit card fraud. As internet banking and digital payments become more prevalent, financial institutions, such as banks, are hindered in their ability to promptly and precisely identify fraudulent transactions. The intricate connections between purchases and cardholders are not always detected by conventional fraud detection technologies. Because of this, the process of identifying fraud is more time-consuming and expensive. In the proposed method, Graph Neural Networks are employed to depict the connections between consumers, transactions, and merchants as connected graphs. This enables the computer to identify suspicious patterns and concealed connections between fraudulent activities. An anomaly detection system that is based on autoencoders is employed to identify any deviations from the norm that may indicate fraudulent activity. These two methods are combined to enhance the system's object detection capabilities and minimize the occurrence of false positives. Banks can promptly identify and prevent fraudulent transfers, despite the fact that the method is intended to operate in real time. Experiments demonstrate that the combination of deep learning and graph-based analysis is significantly more reliable and effective in detecting misconduct in contemporary financial networks.