Research on Credit Risk Prediction Based on Cart Classification Tree

Abstract

The rapid development of domestic Internet technology has made the domestic Internet + industry model a great success. Various traditional industries have produced some new industries after combining Internet technology. Since the beginning of the 21 st century, the connection between the Internet and the financial industry has become increasingly close, which has spawned a new financial model of Internet finance. P2P platform is one of them. Because its lending process is simple and convenient, and there is no strict loan application standard like traditional banks, it has been favored by many small and mediumsized enterprises. Various P2P platforms have mushroomed. However, the rapid development of P2P platforms has also brought security problems that cannot be ignored. This is largely due to the lack of adequate credit evaluation of borrowers before loans or the lack of accuracy of credit evaluation methods. Therefore, the accurate credit evaluation of the relevant lending data of the borrower has become the entry point for reducing the risk of borrowing. Considering that machine learning has been very mature in processing and analyzing data, and has many successful experiences, the CART classification tree model has the advantages of good effect, easy to understand, and less affected by outliers and missing values. It can also find fields that have important warning effects on the risk of borrowing, so this experiment uses the CART classification tree to train the data. Considering that only a single CART classification tree is used to analyze data, there is still room for improvement in analysis accuracy, so this model is optimized to use an Ensemble model to analyze data. The results show that a single CART classification tree model has high accuracy in credit evaluation of borrowers.