The Research On Credit Card Application Scoring Model Based On Data Mining Techniques

After a long market cultivation stage, credit card business in china has entered the rapidly rising development stage. Every Commercial bank increases the intensity of the issuance of credit cards,attracted by the high profits and big market space in credit card business.But high profit follows high risk.It is an urgent issue to do credit scoring effectively for new applicants,which can prevent bad customers and resist credit card risk.So it is meaningful both theoretically and practically to study credit scoring.This paper uses three data mining methods-Decision Tree,logistic regression and neural network and advanced statistic and business inteligent software SAS to buid a credit card application scoring model,based on the customer data in a business bank,which eventually achieves the goal to analyse the complex factors caused credit risk quantitatively.Fistly,we analyse the theory in credit card business and data mining systematically,and then study the extensive use of data mining in every aspect of credit card business.Then we use decision tree,logistic regression and neural network to buid and assess the three different credit scoring models based on relevent data.In the end we aggregate the three models to improve the accuracy and stability,which eventually assess the credit applicants effectively,objectively,accurately and consistently.The final results showed that in the absence of selecting the optimal threshold value, the overall, correct rate of the model is 88.73%, in which 83.44% of "bad customers" and 92.02% of "good customers" are correctly classified.The two main innovation in this paper are as follows:1. Using decision tree,logistic regression and neural network to build and assess different models for credit scoring ;2. building a stable credit card application scoring model using aggragating method integrated the characteristics of the three model .