Study On Credit Rating Model For Business Customer Of Commercial Bank

With the outbreak of financial risk and the rapid development of domestic enterprises in recent years,the whole society is extremely dependent on the commercial banks'credit resources.The commercial banks are the special enterprises.The impact of credit risk(mainly refers to the lack of enterprise customer credit evaluation)is very big on the bank.Due to the restriction of the New Basel Capital Accord,the high risk of enterprise credit consumed commercial bank more capital and reduced its profitability;on the other hand,the lack of standardized financial system is very normal in enterprises,furthermore the asymmetry and opaque information between bank and enterprise strongly impact on the credit judgment for commercial bank to enterprises.Therefore,how to evaluate the customer's credit situation have become the most important part of the development and risk management of commercial banks.In foreign countries,the credit rating,such as ZETA model,decision tree model based on artificial intelligence,K-nearest neighbor classification method,have been developed as a major means of assessing and preventing credit risks.However,this aspect research is still in the relatively backward stage in China,where we use the subjective analysis method or the expert evaluation method.Unfortunately,subjective analysis or expert evaluation method are excessively dependent on the experience of experts.As the asymmetry and opaque information between bank and enterprise is very normal in our country,long-term use of this method will have certain limitations;based on traditional statistical model,because of distortion or uncomprehensive financial information in small enterprises,the authenticity of the use of model will also be greatly limited.For the modern credit risk assessment model and artificial intelligence model,the enterprise scale,historical data accumulation,and credit system construction not fit the enterprise credit rating at this time.In this study,we established a more detailed evaluation system for enterprise credit rating.We used the analytic hierarchy process to select the index system to extract the main component index.We introduced the financial and non-financial information into this index system and optimized the index structure,which provided data support for subsequent model input.On this basis,we used BP neural network,pass configuration,training,testing to complete the self-learning of this model.Using different credit rating(AA~+,A,BBB),we selected the relevant and non-financial data from a commercial bank's head office credit customer to train this model.We also used MATLAB simulation to ensure the reasonable and reliable of data.Finally,other data are used to validate this model.These results indicate that the credit rating of this model has a lower false positive rate,which make it feasible.