The Comparison Of Several Credit Scoring Models

The earnings of commercial banks nowadays are no longer just rely on the interest rate between loan and deposit. They turn to the intermediate business one after another. As a result, the credit card industry has become a fierce competition market although it just developed for a decade. And as the expectation of the liberalization of interest rates, the commercial banks hope to lay a good foundation as much as possible in order to increase bargaining chips for the future competition, but they may ignore the hidden credit crisis.Therefore, the risk management of credit card has become a task of top priority. The traditional method of subjective personal credit scoring no longer meet the needs of contemporary management, we must apply the modern quantitative scoring models and techniques to monitor and manage the risk.This article first introduced the definitions of credit card and its risk, then explain the necessity of credit scoring models on the basis of the analysis of its development, also made a summary of related theoretical studies.This was followed by the introduction of credit score technology development, focusing on the several credit scoring model, such as discriminant analysis, Logistic Regression, Classification Tree, Neural Networks and SVM classification. On the basis of credit card customer sample data, we create these models and compute each result, then we make a comparison in accuracy and stability of these models. At last we get the conclusion that we should use different models under different situations.In the end, after describing the current situation of the domestic credit market risk management, we put forward some proposals to control the risk, and predict the direction of future research.