Research On The Application Of Neural Network To Credit Risk Assessment For Commercial Banks

With the trend of Economic Globalization, the volatility of Financial Markets is intensifying. AS the major financial institutions, the credit of Commercial bank is becoming more serious. AS the main asset business, loan has about 60% of total business. While company as major customers of commercial bank, the credit risk directly related to the vital interests of the bank, is gradually being attention by academic and financial.Assessment as a way or means to guard against credit risk has made great progress in foreign, but the study in our country is still in the initial stage, is lack of systematic, and is backward in the way and means. The study of credit risk for financial institutions should be accelerated as soon as possible in our country, in order to provide more scientific basis for decision. It is meaningful to management and operation scientific and rational for commercial bank.From the perspective of commercial banks, listed company is the main object of the article. First of all, the paper reviews the literatures about the credit risk measurement of other countries and China in this field. Then introduce the related concepts of credit risk of commercial bank. Secondly, the paper introduces the developing of the credit evaluated models briefly and compares them. Basing on this, the paper chose the models which are more suitable to measure the credit risk of our country's listed companies, and explaining the theories of these models detailed.Thirdly, based on research results and experiences in foreign countries and the actual situation in China, the paper initially selected 18 financial ratios to constructed evaluation index system. Proven, some indicators are significantly related. So this article simplifies assessment index system by SPSS. Then the paper chose 102 listed companies which are not special treated and 54 which are special treated as the samples, and design a profit BP-Neural Network model in Matlab.R2010b. After many times of train, we get a profit model.The last part is simulation. In order to test the effect of the model, we need to simulative it. Input the index of these companies which are not use to train. Compare the outputs with targets, we get the result. The correct rate of this model is 86.54%. Can be seen, the model of credit risk assessment on listed companies is effective.