Three Kinds Of Credit Risk Measurement Based On GBDT And Harsanyi Transformation Driven By Them In Internet Finance

In Internet Finance,there is an incomplete information game between Internet loan enterprise and loan applicant.But in practice of Harsanyi transformation,it's difficult to get a new applicat's credit probability distributions.This paper attempts to solve this problem by using statistical learning method,and the main research work includes the following three parts.Frist,an incomplete information Internet loan credit game(3ILCG)model is constructed in this paper.Then the traditional Harsanyi transformation analysis is used for theoretical analysis.And the method of using statistical learning to predict credit risk of loan applicants is proposed to drive the Harsanyi transformation.Second,measuring credit risk of loan applicants by three kinds of statistical learning methods.The definition of credit risk is given in this paper,and measure it by Gradient Boosting Decision Tree(GBDT)model.Then this paper proposed GBDT coupled with Support Vector Machine(SVM-GBDT)model.SVM-GBDT model selects the support vector in SVM as a new training test,it can greatly reduce the size of data while guarantee data information.SVM-GBDT model is used to measure the credit risk,and its result shows that SVM-GBDT model can improve the efficiency 73.72% and guarantee the accuracy than GBDT model.Last the eXtreme Gradient Boosting(XGBoost)model is used to measure the credit risk.And its result shows that XGBoost model can improve the accuracy 0.0107 and efficiency 44.34% than GBDT model.Comparing three methods,it can see that the SVM-GBDT model can be used in a large scale credit data,and the XGBoost model is preferred in general.Third,this paper proposed the Harsanyi transformation driven by credit risk.Based on the characteristics of the data in this paper,the Harsanyi transformation driven by XGBoost is used in 3ILCG.Empirical analysis shows that the accuracy of loan decision made by Internet loan enterprise is 94.8%.Therefore,in the face of a new loan applicant,the Harsanyi transformation driven by credit risk proposed in this paper will help Internet loan enterprise to make scientific loan decision,which can help the enterprise develops steadily and healthily.