Preventing Fraudulent Activities Using Graph Neural Networks

Fraud is a problem found by a lot of people from individual to government institution. Credit card fraud, can cause substantial financial losses. To prevent fraud, it is crucial to develop models that can detect fraudulent activities early by analyzing transaction patterns. Graph Neural Networks, also known as GNNs, helps to deal with this problem. GNNs utilizes the complex relationship in transactions to identify intricate patterns, then label the suspicious patterns as fraud. In this study, we used a GNN-based model for credit card fraud detection that achieves an accuracy of 99.81%. Additionally, we identify the most important features utilized by the GNN model. Our finding demonstrates the effectiveness of GNN in detecting credit fraud and provides valuable insight into feature selection for fraud detection models.