A DQN-based Internet Financial Fraud Transaction Detection Method

Abstract

The anti-fraud issue of Internet finance is a hot research topic in the industry. Aiming at the complex fraud problem of Internet finance, this paper proposes a fraudulent transaction detection method based on Deep Q Learning, and constructs a feasible electronic transaction fraud detection model. Based on reinforcement learning, this method makes the agent learn classification strategies, builds the environment with RFM model, and uses SmoothL1 as the loss function to improve the learning efficiency of the agent. The experiment uses a variety of evaluation metrics to verify the performance. The results demonstrated that the proposed DQN-based fraud detection method in this paper has improved some performance evaluation metrics compared with the traditional method.