Risk Prediction And Evaluation Of P2P Network Loan Platform Based On Random Forest Model

Since 2014,China's P2P online lending industry has been in a strong supervision stage.This can be seen from the current intensive online lending policy.The strengthening of supervision has directly led to the large-scale reshuffle of the industry.The informal and small-scale platforms are gradually withdrawing from the market.With the issuance of the "Notice on Doing a Good Job of P2P Network Lending Risk Special Rectification and Acceptance Work" and the issuance of detailed rules for the acceptance and rectification of various platforms,projects of certain problem platforms have been suspended or ordered to be released.At the same time,financial deleveraging and tightening of funds have led to a run-off,crisis on the platform.Most platforms are facing many crises,such as unable to withdraw cash,unable to operate normally or even bankruptcy.Not only will investors suffer heavy economic losses,but also will have a negative impact on China's social stability and financial order.The risks behind the P2P network lending industry,especially the risks of the platform,are becoming more and more serious,and the risk control of the P2P platform is becoming more and more important.Therefore,it is of great significance to construct a P2P platform risk assessment model to effectively identify platform risks.This paper sorts out the previous research on the risk of P2P online lending platform,and selects a comprehensive indicator that reflects the risk of P2P online lending platform.Firstly,the descriptive analysis of the obtained index data is carried out firstly,then the random forest structure P2P network loan platform risk evaluation model is used and the prediction accuracy of the random forest model is obtained through the confusion matrix,and then the support vector machine and the logistic model of machine learning.In contrast,the analysis yields the optimal model.Finally,the degree of importance of the indicators obtained through random forests is ranked by factor analysis using the factor analysis.The results show that the prediction effect of the random forest model is better than that of the support vector machine and the logistic model.The five indicators,such as volume,average expected rate of return,number of investors,borrowing scales,and full-time bidding,are more important for the risk identification of P2P online lending platforms.Investors can focus on these indicators to evaluate the risk of the platform.The top three platforms in the p2p platform risk assessment obtained by factor analysis are Lujinfu,Niwodai,and Madailicai.