Research On Credit Risk Assessment Of Commercial Banks Based On BP Neural Network

As the main source of income of commercial banks,the income of credit business is ranked first among all businesses.Meanwhile,commercial banks must face the risks of credit business as well.Therefore,credit risk is the most important uncertain factor in the process of bank management.In the process of credit management,the evaluation of credit risk is the most basic and most important link,which has very important theoretical and practical significance.This paper studies the default risk of the credit clients of Chinese commercial banks(this article mainly studies: listed companies).First of all,the credit risk of commercial banks is reviewed,and the causes and characteristics of credit risk are analyzed.The comparison between the traditional evaluation method of credit risk and the evaluation method based on neural network shows that the evaluation method based on neural network is feasible and superior.Then,on the basis of the results of previous studies at home and abroad,establish the credit risk evaluation index system,the actual situation of our country including: profit debtpaying ability ability,development ability,operation ability,cash flow,enterprise credit and enterprise competition ability of 7 first level indexes and 25 level two indexes.Then the SPSS is used to analyze the correlation,and the financial indicators of high degree of relevance are gathered into one class.By using factor analysis,the most representative indexes were selected in each class.After screening,the index was simplified to 19.Then,based on the simplified index system,a three level BP neural network model is constructed,and Bayesian regularization and momentum gradient descent method are applied to solve the traditional BP artificial mental over fitting phenomenon.Finally,the annual data of the randomly selected 101 non ST companies and 25 ST companies in the Shanghai and Shen zhen two cities are taken as sample data.And the Matlab.R2010 b is used to train and test the sample data.The experimental results show that the accuracy of the credit risk assessment model based on BP neural network is up to 95%.Therefore,the above model can realize the accurate evaluation of the credit risk of the enterprise customers,so as to improve the credit risk assessment of the commercial banks in China.