Title:detectGNN: Harnessing Graph Neural Networks for Enhanced Fraud Detection in Credit Card Transactions

Abstract:Credit card fraud is a major issue nowadays, costing huge money and affecting trust in financial systems. Traditional fraud detection methods often fail to detect advanced and growing fraud techniques. This study focuses on using Graph Neural Networks (GNNs) to improve fraud detection by analyzing transactions as a network of connected data points, such as accounts, traders, and devices. The proposed "detectGNN" model uses advanced features like time-based patterns and dynamic updates to expose hidden fraud and improve detection accuracy. Tests show that GNNs perform better than traditional methods in finding complex and multi-layered fraud. The model also addresses real-time processing, data imbalance, and privacy concerns, making it practical for real-world use. This research shows that GNNs can provide a powerful, accurate, and a scalable solution for detecting fraud. Future work will focus on making the models easier to understand, privacy-friendly, and adaptable to new types of fraud, ensuring safer financial transactions in the digital world.