Research On Credit Risk Assessment Of P2P Online Loan Borrowers Based On Multi Model Fusion

As a new financial model,P2P online loan combines Internet with traditional credit.It develops rapidly with effective financing efficiency,convenient transaction process and wide coverage.However,due to the lack of a perfect credit supervision system and the problem of information asymmetry,in recent years,P2 P platform thunderbolt phenomenon occurs frequently,which seriously damages the interests of investors and leads to the trust crisis in the online lending industry.Credit risk is an important reason for the frequent occurrence of platform problems,so the establishment of a scientific and effective credit risk assessment system is one of the important measures to promote the sustainable and healthy development of P2 P online lending platform.In view of the above problems,this paper introduces random forest algorithm in the index selection of credit evaluation system.The importance of each feature is calculated by random forest algorithm,and then sorted in descending order according to the importance of each feature.The first 64 features are reduced to 46 features,which greatly optimizes the feature set of the original data set and improves the index composition of the credit evaluation system.At the same time,in order to improve the perform ance of the credit risk assessment system,this paper proposes a P2 P borrower credit risk assessment system based on multi model fusion.According to the idea of voting method,it fuses logistic regression classification algorithm,random forest classifica tion algorithm and lightgbm classification algorithm according to the principle that the minority is subordinate to the majority,corrects the deviation problem of single model prediction,and selects the accuracy rate,random forest classification algorit hm and lightgbm classification algorithm Recall rate,F1 score and AUC were compared with three single models on lending Club dataset.The multi model fusion algorithm proposed in this paper has the best prediction performance in four evaluation indexes,i ncluding accuracy rate of 0.9961,recall rate of 0.9765,F1 score of 0.9862 and AUC value of0.9864,which fully verifies the feasibility and robustness of the multi model fusion algorithm based on voting method,which can more scientifically evaluate the credit risk of borrowers and reduce the credit risk of borrowers and Internet users The risk of online lending platform in the process of lending can promote the healthy development of online lending platform.