GNN-DRIVEN DETECTION OF ANOMALOUS TRANSACTIONS IN E-COMMERCE SYSTEMS
HaoYu Wu, JiaYi Wang · Journal of trends in applied science and advanced technologies. · 2025
The exponential growth of e-commerce platforms has transformed global trade, enabling seamless digital transactions. However, this expansion has also led to an increase in fraudulent activities, including fake transactions, money laundering, and synthetic account fraud. Traditional fraud detection systems, which rely on predefined rules or supervised learning models, struggle to adapt to evolving fraudulent tactics. This study proposes a graph neural network (GNN)-driven anomaly detection framework to improve fraud detection in e-commerce systems by leveraging the inherent graph structure of online transactions. The proposed approach models e-commerce transactions as a heterogeneous transaction graph, where nodes represent users, merchants, and transaction records, while edges encode relationships such as purchase behavior, payment connections, and review activity. The framework integrates graph convolutional networks (GCN) and graph attention networks (GAT) for spatial anomaly detection, combined with temporal graph networks to track transaction sequence patterns. Unlike traditional methods, this approach captures both structural transaction dependencies and time-based anomalies, enabling the detection of coordinated fraud schemes. Extensive experiments on real-world e-commerce transaction datasets demonstrate that the proposed model outperforms conventional fraud detection techniques, achieving a higher detection accuracy and a significantly lower false positive rate. The results highlight the effectiveness of graph-based learning in identifying complex fraud rings, transaction laundering, and fraudulent refund behaviors. This research underscores the importance of GNN-powered fraud detection in enhancing e-commerce security, providing an adaptive and scalable solution for modern digital marketplaces.