Research On Applications Of Data Mining Techniques In Credit Card Approval

In recent years, credit card business developed rapidly since the raise of income and the change of consumer's habit. The credit business becomes the core business which can cause plenty of customer and considerable profit for banks. The credit card market in Chinese competes intensely day by day, But highly competition causes limits of people who apply for credit cards in commercial banks tend to loosely and they neglect risk management and control.This thesis focuses on the research on credit scoring of those applying for credit cards, which belongs to the crediting link within the risk management of credit cards. The tool for data mining is Clementine software, SPSS core mining product. According to the industry standard of CRISP-DM provided by SPSS Co., the research sets up the credit scoring model with the five steps: business understanding, data understanding, data preparation, modeling, Evaluation and Deployment. After carried exploration analysis and preprocessing on data, three methods for setting up models are utilized in this thesis: the first one is decision tree, the second is neural net and the third is logistic regression. And taking result, mis-classification rate, efficiency and theoretics as general evaluation criteria, compare the three methods for setting up models, and acquire the conclusion that logistic regression model is the optimum scoring model. By means of logistic regression model, the establishment of credit scoring on applicants of credit cards can help commercial banks make an objective, coincident and quick judgement. This can impels the bank's credit card business develop well and enhances commercial banks'ability of risk guard.