Research On Credit Risk Prediction Model Based On Machine Learning Technology

Internet consumer finance is the Internet financial companies to meet the individual consumers' demand for goods and services provided by the consumer demand for small loans and amortization of credit activities.Compared with the traditional consumer finance,Internet consumer finance is a convenient and efficient service,significantly reduce transaction costs,also covering a wider range of consumers.With the continuous development of economic,people's consumption concept upgrade,Internet consumer finance has gradually been recognized by more and more consumers.In the face of tens of thousands or even one hundred thousand of the users to apply for loans,it's important to predict the user's credit risk through the Internet and computer technology.In this paper,we studied small loans for Internet financial consumer applications,to explore the development of machine learning techniques in this field and the its practical application.Logistic regression model and GBDT model(gradient boosting decision tree model)is popular in the field of credit risk assessment.As well as the mainstream model of performance evaluation index.In the experiment stage,this paper uses machine learning technology of user modeling and analyzing network credit platform,the variable selection features using Information Value statistics,use woe(weight of evidence)encoding the ordinary value of variables,with the ability to distinguish,enhance the variables on the two types of users.Finally,using Logistic regression model and GBDT model(gradient boosting decision tree model)to apply for a loan credit risk prediction model of user modeling,and using AUC as the specific indicators of model performance verification,further optimization of the model.Finally,discussed the development of credit risk model is summarized and prospected.