Fraud Detection in Financial Transactions Using Graph Neural Networks

Fraud detection in financial transactions remains a critical challenge due to the complex, interconnected nature of modern fraudulent schemes. Traditional machine learning approaches analyse transactions in isolation and fail to capture relational patterns indicative of coordinated fraud rings. This paper presents a comprehensive full-stack system combining a Graph Neural Network (GNN) backend with a modern React.js frontend for real-time fraud detection and visualisation. The backend employs Graph Attention Networks (GAT) and Random Forest ensemble methods to model transactional relationships and identify anomalous patterns. The frontend provides an interactive dashboard enabling users to submit transactions, receive real-time predictions with confidence scores, visualise analytics through dynamic charts, and manage transaction histories. Deployed using Flask and containerised with Docker, the system achieves scalability and robustness for production environments. The modular architecture supports seamless integration with banking and fintech platforms, demonstrating the efficacy of graph-based learning combined with intuitive user interfaces in advancing financial security.