Study On Credit Rating Of Small Enterprises Based On Optimal Index Combination And Optimal Weight

Credit is a loan activity based on repayment of principal and interest.Credit risk,also known as default risk,refers to the possibility that the borrower or the issuer is unwilling or unable to fulfill the contract agreement for the subjective or objective reasons,resulting in the loss of the bank and other financial institutions,institutional investors and individual investors.The risk of default is the most important risk of banks.It is the main cause of bad debts,deficits and even bankruptcy of banks.The credit rating is used to identify the default risk of a loan or a corporate bond and provide an important basis for the decision of the bank loan and the company's bond investment,by the relationship between the rating model and the default state.In the big data environment,the key to credit rating is to get the relationship between the index data and customer default status.First,how to select a optimal index combination with the strongest ability to identify credit risk.Second,how to determine a optimal weight vector with the strongest ability of default identification in the infinite group weight vector.Third,how to determine the "default misjudgment non-default" and "non-default misjudged default" two types of error costs,which ensures the maximum accuracy rate of credit rating and minimize the error of "default misjudgment non-default".The greatest credit rating model based on optimal index combination and optimal weight vector can be applied to the credit rating of different subjects,such as small enterprises,listed companies,farmers,merchants,personal loans and bonds.It provides decision-making reference for banks to issue loans,enterprises to issue bonds,guarantee corporation to undertake credit guarantee and personal investment.This study consists of four parts: the first chapter is the introduction,the analysis of the background and significance of the research,the status of the research on the credit evaluation of small enterprises and the existing problems.The second chapter is to build a credit rating model based on optimal index combination and optimal weight vector.The third chapter takes small enterprises credit rating as an example to conduct an empirical study.Establish the optimal credit rating system for small enterprises in China.The fourth chapter summarizes the conclusions,innovations and characteristics of this study.The main work of this study is:First,delete redundant indicators.In the index system of m indexes,delete each index of redundant index with correlation coefficient greater than 0.8,and build 2 index system composed of m-1 indexes.Calculate this 2 index systems' F value.Retain the index system with larger F value.Not only avoid the information duplication in the index system,but also avoid error deleting the index with greater breach of default.It has changed the shortcomings of existing research in determining the optimal index system,which ignore the actual discriminating ability of the customer's default risk.Second,set up the optimal credit rating index system.Using the Lasso-logistic regression model,minimize the credit risk rating deviance,and build the optimal index combination with the weight of index not equal to 0.Ensure the credit rating index system has the maximum default discrimination.It improves the existing research that the optimal index system just was selected from the limited number of subjective indicators.As a matter of fact,select from limited subjective index combinations only get local optimal solution,rather than the global optimal solution.Taking the credit rating of small enterprises as an example,we select the optimal index system with 48 indexes.Third,determine the "default misjudgment non-default" and "non-default misjudged default" error costs.With the goal of maximum accuracy rate of credit rating and minimize the two types of errors cost,as well as the error cost of "default misjudgment non-default" larger than "non-default misjudged default",set up the optimal two types of error cost.Avoid the existing research just taking the maximum credit rating accuracy as the standard to reverse the optimal weight,and the equal treatment of two kinds of error cost,which will cause more "default misjudgment non-default" error and excessive credit loss.The empirical study of small enterprises shows that the optimal ratio of two error cost is 6.Fourth,build the credit rating equation with optimal weight vector.In the pursuit of the minimum total error costs and the "default misjudgment non-default" error with lager weight,set up the goal of maximum accuracy rate of credit rating to reverse the optimal weight.It ensures that the credit rating equation with optimal weight can distinguish the default and non defaulting customers significantly.