Research on Risk Evaluation Algorithm of Private Lending Based on ROC Decision Tree

Abstract

In order to enhance the risk control of private loans, a risk assessment technology of network loan projects based on multi-feature information and a risk-sensitive decision tree is provided. In this method, the credit conditions of the lender in the customer information submitted in the online loan system is first established and then combined with the lender's historical loan records, personal characteristics, and enterprise characteristics. The analysis method with the maximization under the ROC curve and the minimum mutual information are used to select the information set most relevant to the project risk analysis and with the minimum redundancy. Finally, the risk evaluation of loan enterprises is carried out by using the perfect risk-sensitive decision tree method. The test result shows that this method has better performance in private loan risk assessment compared with the existing methods.