Personal Credit Evaluation Model Research Based On The Signal Noise Difference

With the rapid development of China’s economy, change of residents’ consumption concept, and the positive support of credit policy, the personal credit business of the commercial banks in China is developing,which is an increasingly large scale.In this rapid development process, the expansion of credit transactions in bank credit risk is more and more complex,the commercial bank credit risk management becomes a key part in risk management system.In order to strengthen the ability of commercial banks to resist the risk, reduce their potential losses, the personal credit risk assessment and prediction of commercial bank has attracted more and more attention.Under this background, this paper chooses the personal credit risk assessment of commercial banks to study, explores the individual credit risk prediction and assessment methods of commercial banks.Not only can effectively reduce the cost of commercial bank credit assessment and potential losses, but also have a positive impact for China’s indirect financing scale expansion, healthy and continuous development of financial market,it has great significance on theory and practice.Most of the literature using credit score card model, namely through the screening of the personal characteristics index with high information to build model with the aid of certain modeling analysis method.Among them, how to extract variables with most value and the highest information from individual characteristic variables provided by the customers, how to code and group the variables with a scientific and reasonable way, these have become the key problems which commercial bank encountered in the personal credit risk assessment.Based on the strong interest to this topic as well as its application value,this article selects customer data from a bank of German and a bank of Australia as sample data,using SAS statistical software, studying the personal credit risk assessment and prediction under different methods.Signal noise difference methods is introduced in this article,it was compared with the traditional T test method,which embodied the superiority of signal noise difference method in the variable screening through empirical test. In addition, in terms of encoding of variables, this article provides two different coding methods:the traditional step by step from more to less group gradually merge and from less to more based on the signal noise difference method, combined with the SAS program to empirical test the effectiveness of two methods. It provides another worthy method in credit risk assessment for commercial banks.In terms of prediction model building,the credit score card model based on Logistic regression analysis and the default probability prediction model based on fixed the naive bayesian classification were adopted in this paper. From the result of empirical test,we know for a bank of German and a bank of Australia, the model using Logistic regression analysis based on noise difference variable encoding has almost the same precision with the model based on traditional variable encoding method. In addition, on the whole, the model using Logistic regression analysis has the higher forecast accuracy than the model using fixed the naive bayesian classification. These show that the signal noise difference method has a high reference and application value in screening and coding of variables, and it is a better choice to build a credit rating model using Logistic regression analysis method, and fixed the naive bayesian classification is also a worthy of reference method due to the advantages of simple.