Research On The Application Of XGBoost Algorithm In Online Lending Credit Risk Assessment

In recent years,with the continuous progress of Internet technology and the increasing enrichment of Internet products,China has become the country with the largest number of Internet users.On the other hand,commercial Banks are unable or unwilling to provide lending services for small and micro businesses and individuals due to the influence of operating costs,technical level,"information asymmetry" and other factors.In addition,the service capacity of private lending is not enough to meet the needs of all small and micro businesses and individuals.In addition,some private lending companies have high interest rates and opaque operation,which further aggravate the financing difficulties of small and micro businesses and individuals.The emergence of online lending has greatly alleviated this contradiction and thus achieved rapid development.However,the rapid growth of the online lending industry cannot conceal risks,and more and more online lending platforms have been forced to close down or transform due to poor management,which has aroused widespread concern in the society.The occurrence of risks,not only the online lending platform itself does not pay enough attention to credit risk and the lack of risk management technology,but also the borrowers are unable to repay or unwilling to repay the loans,and some borrowers deliberately falsify their identities in order to defraud the loans.It is of great significance to establish an effective credit risk management and evaluation system for online lending,and to warn and prevent the potential credit default risk of borrowers in time.The purpose of this paper is to explore an efficient credit risk evaluation model for online lending,so as to provide some references for the credit default risk assessment of borrowers and the healthy development of online lending industry in China in the future.This paper studies the credit risk assessment of borrowers from the perspective of online lending lenders and online lending platforms,conducts research and analysis of online lending and credit risk assessment through literature research method,and makes a comparative analysis of traditional lending and online lending credit risk assessment methods using comparative analysis method,and proposes a credit risk assessment model of online lending based on XGBoost algorithm.In the model preparation stage,feature generation,feature processing and data description statistics are respectively carried out,and model assumptions are proposed.Then model training and feature analysis are carried out,and a borrower default prediction model is preliminarily constructed.In the model comparison stage,this model is compared with the commonly used classification models such as Logistic regression,decision tree,random forest and GBDT,etc.,and it is concluded that XGBoost model is more suitable for the measurement of online lending risk assessment.In the model optimization stage,it is mainly realized by adjusting the parameters,and the model is evaluated and analyzed by using the confusion matrix,ROC curve,AUC value and other indicators.Finally,an efficient online lending risk assessment model is obtained.At present,the online lending industry is still in the process of rapid development,the sample data selected in this paper cannot cover all users,so the credit risk assessment model has some limitations in application,but it has practical significance in the method and practice of online lending credit risk assessment.At the end of the paper,some suggestions based on XGBoost algorithm for online lending credit risk assessment research are put forward.