The Study On Credit Card Default Of The Bank Based On Data Mining

With the globalization of finance and the reform and development of China's financial system,credit card business,as one of the fastest-growing product lines in China's personal financial services market,has become a breakthrough as the rapid growth of bank profits.However,as an unsecured credit facility,credit cards also pose a huge risk behind the high returns.Therefore,how to effectively use the data generated by credit card holder and use data mining to find useful information for controlling risks has become one of the issues that banks have focused on.The paper uses the personal data of bank card holder and the credit card spending and repayment history for six consecutive months as the input variable,and uses the default payment next month as the output variable to research credit card default.After briefly introducing the data mining technology,SVM,kNN,Decision Tree,Random Forest and LightGBM algorithms are used and combined the sample data to establish the classification model respectively.After comparing the confusion matrix and three related classification accuracy evaluation indicators,we find that the overall ranking of the five algorithms from merit to bad is as follows: LightGBM,Random Forest,Decision Tree,SVM,kNN;In terms of the category of default,the ranking is: LightGBM,Random Forest,Decision Tree,kNN,SVM.The LightGBM model has the best classification effect among all the models constructed in this paper.In addition,this paper also uses the decision tree rule extraction method to dig out several key features of the data set.The consistency with the objective facts proves that the model is valuable for banks to prevent and manage credit card default risk.