Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Июль 17, 2024

Abstract Fraud seriously threatens individual interests and social stability, so fraud detection has attracted much attention in recent years. In scenarios such as media, fraudsters typically hide among numerous benign users, constituting only a small minority often forming “small gangs”. Due to the scarcity of fraudsters, conventional graph neural network might overlook or obscure critical information, leading insufficient representation characteristics. To address these issues, tran-smote on graphs (GTS) method for is proposed by this study. Structural features each type node are deeply mined using subgraph extractor, integrated with attribute transformer technology, node’s information enriched, thereby addressing issue inadequate feature representation. Additionally, approach involves setting embedding space generate new nodes representing classes, an edge generator used provide relevant connection nodes, alleviating class imbalance problem. The results from experiments two real datasets demonstrate that GTS, performs better than current state-of-the-art baseline.

Язык: Английский