Research On The Prediction And Analysis Of Customers' Credit Card Default Of A Bank

The change of social consumption mode makes more and more commercial banks compete fiercely in credit card business.However,as a kind of credit certificate issued by commercial banks to individuals or units with good reputation,credit card brings profits to banks,but also hides some default risks.Reducing the losses caused by the default of credit card customers is the common goal to improve the profits of credit card business of major commercial banks.For A bank,its customers' relationship management information system records a large number of credit card customers' information data.These information data are some effective activity information of credit card customers.These information data can effectively send default warnings to the bank in a period of time before the default of credit card customers.If the bank can effectively use these information data,it can take appropriate measures to reduce the losses caused by credit card customers' default after early warnings.Therefore,the key to solve the problem of A bank's credit card customer default is to effectively use these information data to establish an accurate credit card customer default prediction model.This paper first sorts out the relevant research theories and results at home and abroad,and combines data such as customer credit information,customer application information,arrears history information and consumption history data obtained from A Bank.These data are cleaned and transformed,unbalanced data processing and variable dimension reduction processing.Secondly,Logistic Regression,Decision Tree and BP Neural Network are combined with preprocessed data to establish classification models.By comprehensively examining the performance classification accuracy,default classification accuracy,overall classification accuracy and AUC value of the three models.It can be concluded that the predictive performance of BP Neural Network is better than the Logistic Regression model and Decision Tree model,and BP Neural Network is suitable for A Bank to perform credit card customers' default management.In addition,this paper also conducted a multivariate correlation and regression analysis on the default amount of bank's credit card customers,and found the influencing factors of the default amount of bank's credit card default customers.Finally,this paper puts forward three suggestions to manage the credit card business of A Bank based on the above research findings.