The Study And Realization Of Credit Card Approval Model Based On Logistic Regression

Credit card audit scoring is a necessary means of commercial bank risk control. The primary basis of carrying out the credit card application score is the in-depth data mining for the consumer data. This paper has a research on the principle and some main realization technology based on the credit card application scoring system of the binary logistic regression model. The research work relies on an actual project on the credit card scoring system of a domestic commercial bank and a credit card scoring prototype of a foreign bank.This paper first studies the concepts of the generalized linear regression model and the discrete dependent model. And then it gives the modeling process of the binary logistic regression model and shows the detailed analysis explanations of an instance.The author obtains the financial history data and research environment by the cooperation unit and achieves the expected objective by the cooperation of all the team members. The author shoulders the specific work as following:1. Data integration, data exploring, data cleaning, variables selection and transforming. In one word, the designing, programming and debugging of SAS codes of the whole ETL.2. Based on the real financial history data of binary logistic regression prototype research, modeling, model experiment and model test.This paper shows that, the personal credit rating evaluation binary logistic regression model which the author participated in it has some outstanding advantages including the complete theory, the robust model and the strong interpretability of the model coefficients. And it is a good and applicable model by the inspection and certification of the practice. The SAS codes of ETL module have the advantage of the clear logic, compact structure and reusable macros.