The Research On Risk Rating Model Of Applying For Credit Card For Commercial Bank

There are credit risks being consistent across all of the four steps of credit card’s life cycle: the examining and approving of credit card’s application, the management of credit card account, anti-fraud and overdue receivables’collection and write-offs. What’s more, there should be different credit risk rating methods for each step. As far as the first step, the application’s approval or rejection is based on the good or bad credit rating result given from the credit scoring model. After the research and comparison for a variety of credit scoring methods, this article consider that because of less limitation for sample data, and both stable and good effect on predictions, logistic regression has a good adaptability, especially for classification problems, such as credit rating. However, the model built through this method could not recognize and handle the correlation among its independent variables, and so that may cause adverse effects on goodness of fit and capacity of classification forecasting of the model.Therefore, for the credit scoring model built through logistic regression methods, this article first proposed to test the correlativity and its degree among the independent variables which are included by using contingency table analysis; then extract interaction terms from these correlation which determined after practical significance consideration of the interaction among them simultaneously. The results of the empirical analysis shows that substitute interaction terms for a part of single independent variables within the general logistic regression model, could not only make the predicted value of the model to be more approximately to the sample data, but also improve the prediction accuracy of good and bad customer’s classification. Beside that, during the building process of logistic regression model that contains interaction terms, the sample consisted of equal numbers of good and bad consumers, which means their ratio is1:1, should be the optimum choice. Moreover, an appropriate standard value should be set in order to help the model judging applicants’credit rating, and it must be because that the direct result calculated by logistic regression model is the predicted value of their default probability. Hence, if adopting the improved method proposed by this article when building a logistic regression model for applicants’credit scoring, commercial bank would gain more earnings owe to issuing more credit cards to the superior customers, and, furthermore, enhance the early warning capability of dangerous customers so as to the risk aversion. And finally help the bank to obtain both of the benefits and security.