Anomaly Detection in E-Commerce Platforms via Graph Neural Networks

The rapid expansion of e-commerce platforms has introduced significant challenges in fraud detection, including fake reviews, payment fraud, account takeovers, and product listing scams. Traditional fraud detection methods, such as rule-based systems and supervised learning classifiers, struggle to detect sophisticated fraudulent activities that evolve over time. This study proposes a graph neural network (GNN)-based anomaly detection framework to enhance fraud detection in e-commerce platforms by leveraging the graph-structured nature of user interactions, transactions, and review networks. The proposed model constructs an e-commerce interaction graph, where nodes represent users, products, and transactions, while edges capture relationships such as purchases, reviews, and payment flows. The framework utilizes graph convolutional networks (GCN) and graph attention networks (GAT) to learn spatial dependencies within the transaction network, combined with gated recurrent units (GRU) to model temporal fraud patterns. By integrating spatial and temporal learning, the model can identify suspicious user behaviors, fraudulent transactions, and fake product listings with high accuracy. Experiments conducted on real-world e-commerce datasets demonstrate that the GNN-based model outperforms traditional fraud detection approaches in terms of F1-score, precision, recall, and false positive rate reduction. The framework successfully detects anomalous activities with an F1-score of 0.91, significantly improving fraud detection in large-scale e-commerce environments. These results highlight the potential of graph-based deep learning in securing online marketplaces against fraudulent activities.