Application Of Discriminant Analysis In Credit Risk Management Of Commercial Banks

Credit risk is the main risk of domestic commercial banks, and the level of credit risk management directly determines banks’ profitability and sound operation capability. How to manage Credit risk scientifically is the core of domestic commercial banks risk management. For credit risk management, foreign advanced large banks have taken quite a few attempts, and domestic banks can learn from foreign banks combining with the actual situation of the domestic banks.As summary of the advanced banks’ risk management and regulatory. Basel Accord offers recommendations about credit risk management, and how to accurately predict defaults on credit risk management is the key point Forecast for breach of contract based on historical data, and people need to construct reasonable risk management models to quantitatively analyze cus-tomer information. History of advanced foreign bank credit’ research on risk management model has been quite Long, and based on long-term accumula-tion of data, more and more accurate models have been taken into practice. Overall, the development of risk model has gone through three stages-expert judgment method, statistical model analysis and modern risk management model. For domestic banks, data management is just a beginning, and the valid data is very difficult to obtain. Even in large banks, the comprehensive data storage is also from recent year, and it leads to the unavailable of modern model. Now, domestic bank is mainly making research on statistical models.The article first introduces Basel Accord, which is the main basis of the modern banking risk management. Under the background of economic glob-alization, to learn uniform standards of the risks management is necessary prerequisite when participating in international competition, but focus of the article is not in it, so the article will just make a general talk about the devel-opment of Basel, with the purpose to let everyone know banks’ comprehensive risk management. The article focuses on credit risk management, and intro-duces the definition of credit risk management of commercial banks, the types of credit risk, the classification, and the credit risk management models. The third chapter introduces a classic statistical model of commercial banks risk management-discriminant analysis model. Discriminant analysis uses the ex-isting data to build classification criteria, and then determine the classification of new samples. The commonly used method is Fisher discriminant analysis, which projects multivariate data onto a one-dimensional space, and you can visually see the concentration of the data. Fisher discriminant analysis has been applied in many fields, such as biology, medicine and so on. But Fisher discriminant analysis makes a poor performance in high-dimensional data, in which the sample covariance matrix is singular and its inverse is not well de-fined the covariance matrix is irreversible. The accumulation of noise terms will make the Fisher discriminant analysis perform poorly. Domestic banks are lack of effective data, especially default sample data. So it’s possible that the number of feather variances of samples is more than the number of samples, called "small samples", which we call it high-dimensional. For that, we supply two methods-stepwise discriminant analysis and FAIR model, which select the key variables to reduce the sample dimension. After the introduction of theory, the article will make empirical analysis of the actual data of a bank, and draw conclusions.