The Study On Credit Risk Management Of Commercial Bank Based On Data Mining

The finance is the economics arterys of a country development and a core of the economy.Risk management is a core part of financial management and the credit risk management is the most important part that banks have to face.Moreover,with the expansion of credit transactions,the credit risk faced by commercial banks has becoming more complex.How to measure and manage credit risk is an theme concerning the commercial bank managenment.With applying data mining,the paper examines the important part of the risk management,that is,the risk measurement and management of the enterprises with loans,from the perspective of commercial banks,with a desire to offer the technology and methods for our commercial banks.The paper makes application of data mining in credit risk management as the study object.Firstly, theory of credit risk management and data mining technique are introduced and then we study the use of data mining in banks.Secondly,reaearch assessment process of credit risk is constructed.Then,the paper set a system of credit risk assessment which contains five aspects including 25 index.Finally,48 special treated listed companies from 2006 to 2008 and their 48 corresponding listed companies which are informal financial conditions are selected as the comparative samples.With the help of companies'public information on financial data,the paper conducted the K-S to test if the indexes follow the Nomal Distribution,and ues T test,non-parametric test and factor analysis as a means to reduce the number of the variables.Then, contrasted by the SAS-SEMMA framework,the data mining soft ware SAS/EM is used to the realization of application of data mining in credit risk assessment by Logistic Regression,Decision Tree and Neural network approachesFrom the research, credit risk assessment model constructed by data mining has great practical value.Comparing effects of three model assessment,prediction error fraction shows that Neural network's is least,Logistic Regression's is maximum,but Decision Tree's is in the middle.But the best overall effect is Decision Tree model.Further,the model is optimized to combine Decision Tree with Neural network.The research result shows that stability and accuracy of integrated model are improved greatly and obtains better assessment results.