Credit Fraud Detection Strategy Based on Graph Neural Network

Abstract

: In this big data-driven era, the digital characteristics of financial credit are constantly undergoing strengthening, the financial relationship network is becoming increasingly sophisticated, and the forms of credit fraud faced are becoming more severe, which poses a serious threat to the security of financial market transactions and the stable development of the economy. Based on the present situation, this article discusses a categorized approach to detection, starting with the satisfaction of functional requirements. Specifically, it explores detection methods based on graph neural networks from both the perspectives of non-functional and functional requirements, and then clarifies their strengths and weaknesses. Furthermore, the article explores detection methods that combine LLM and graph neural networks. Finally, this article introduces commonly used datasets and proposes several feasible solutions to their limitations, such as data imbalance and privacy protection. In the context of financial credit fraud based on graph neural networks, the comprehensive application of a variety of detection strategies, assisted by LLM and federated learning, will further enhance the robustness and accuracy of financial fraud detection, safeguard the stable operation of financial markets, and promote sustainable economic development.