Risk Prediction of P2P Credit Loans Overdue Based on Gradient Boosting Machine Model

With the gradual integration of online lending into our lives, the credit risk issues it brings have also attracted attention. Credit risk issues not only make the lending platform significantly increase financing costs but also affect the average yield of lending users. Therefore, we collected truthful user data from a P2P company, which counted 3,000 P2P user information while including 446 feature elements, which also conducted a lending user overdue risk prediction study on the collected data. The research content is mainly to explore and clean the data, dig out the significant factors affecting the users’ repayment overdue, modeling the late repayment rate of P2P credit users by data analysis method, and finally ROC curve to evaluate the model performance. The research provides more objective and comprehensive analysis and prediction of the late repayment rate of credit users in the data. The experimental results show that the GBM model can be applied to credit late repayment prediction and avoid the limitations of some models. According to the above mining results, P2P enterprises can be more reasonable and efficient to carry out risk management, identify high-risk credit loans. So that enterprises can be more sustainable and steady development.