TMR-GGNN: Credit Card Fraud Detection based on Time-Aware Multi-Relational Guided Graph Neural Network
Researchers have developed a new framework called TMR-GGNN to combat credit card fraud. This model addresses challenges like imbalanced data and evolving fraud patterns by constructing a dynamic, multi-relational graph that considers interactions between customers, merchants, devices, and IPs over time. It incorporates a time-aware attention mechanism and a contrastive learning decoder to better identify rare fraud cases and reduce false negatives, utilizing a composite loss function that combines contrastive loss with Focal Loss. AI
IMPACT This research introduces a novel graph neural network approach that could improve the accuracy and robustness of credit card fraud detection systems.