Research On Internet Consumption Credit Risk Control Based On Boosting Method

With the rapid growing of the scale of Internet consumption credit lending,the risk control ability will be a decisive factor for whether an institution can maintain a solid foothold in the development of large waves.The construction and evaluation of the risk control model under the support of an effective index system and machine learning algorithms,which improve the model adaptability and accuracy,is an important direction for exploring the Internet credit risk control.This paper conducts in-depth research on the issue of Internet consumer credit risk control,expounds its development status and highlights the importance of risk control model.We analyze the index system of Internet consumer credit risk control under big data and establish an indicator analysis process,by dividing the indicators into six directions and four dimensions.Using the method of text mining to convert text-based indicators into TDM matrix for hierarchical clustering,we convert the data which could not be used originally into derivative indicators,so as to support the risk control model effectively,and enhance the correctness of the model for users' risk.The application of AdaBoost,GBDT,and XGBoost algorithms in the Boosting method to the risk control model is mainly studied.And using real Internet consumer credit data for empirical analysis,we have compared them with random forest algorithm.We establish risk control models under four different algorithms,respectively,to predict the risk degree of the borrower,and finally evaluate the effect of each algorithm.The effectiveness of the Boosting method in the risk control is proved and important indicator is obtained,which can provide reference for institutions to establish risk control models of Internet consumer credit.