Small Business Credit Rating Model Based On The Default Discrimination

Credit is lending business on the condition of debt service.Credit rating aims to reveal the default risk and determine the loss rate by mining the relationship between rating data and default risk.The nature of credit rating is default discrimination.A credit rating system without significant default discrimination is not an effective rating system.The financial crisis in 2008 was mainly due to the failure of the rating system's ability to discriminate default risk.Small businesses are an important part of the national economy.However,their unreliable credit information and irregular management make the credit rating difficult,what's more,banks are hampered by credit risk management and adopt small loans or even no loans to small businesses,which lead to difficult for small businesses to finance,and further restrict the development of small businesses.Therefore,how to build a credit rating model for small businesses to help ease the financing difficulties needs to be resolved.This study uses the default discrimination as the standard to construct the small business credit rating model,which mainly includes three research contents:the model of constructing credit rating indicator system based on the maximum default discrimination of indicator group,the credit grade division model based on the maximum discriminatory ability between grades,and incremental credit evaluation model based on the default discrimination of new sample.The three research contents are closely related and in-depth.The main works and innovations of this thesis are as follows:1.Established the credit rating indicator system model based on the maximum default discrimination of indicator group,and established the credit rating indicator system for small businesses.Through correlation analysis and K-S test statistic D value,the indicator reflecting information duplication was removed.From the generating indicator group by iteratively deleting indicators,the indicator group with the largest D value was selected as the final indicator system,which ensures the indicator system has the most significant ability to discriminate default risk.The model enriches the method of establishing the credit rating indicator system.Using the 3 045 small businesses from a Chinese bank as empirical sample,the study finally builds the small business credit rating indicator system with 21 indicators,including macroscopic factors such as per capita disposable income of urban residents,personal factors such as living conditions and working time,mortgage guarantee situation,etc.The study shows that:(1)Although the single indicator has the maximum discriminatory ability,the indicator system they compose may not have the most significant ability to discriminate default risk.(2)Although the single indicator seems to be very good and popular,the indicator system they compose may not have the significant ability to discriminate default risk.(3)The number of indicators in the credit rating indicator system is not the more the better.2.Established the credit grade division model that meets the dual criteria of the credit grades should have the maximum discrimination and the higher the credit rating,the lower the loss rate,and computed the loss rate of every credit grade.Establishing the credit grade division model by taking the maximum difference between the cumulative frequency of non-defaulting customers and the cumulative frequency of defaulted customers as the objective function and taking the higher credit rating with the lower loss rate as the main constraint,which ensures the divided credit grades meet the dual criteria of the higher the credit rating,the lower the loss rate and the credit grades shouldhave the maximum discrimination,and calculates the loss rate of every credit grades.Using the 3 045 small businesses from a Chinese bank for empirical study,the empirical results show that the divided credit grades meet the criteria that the higher the credit rating,the lower the loss rate,and the credit grades have significant discrimination.3.Established the incremental Bayesian network credit evaluation model based on the default discrimination of new sample.By building a Bayesian network credit evaluation model based on the old sample,and then updating the current Bayesian network credit evaluation model with new and old sample,this paper established the incremental credit evaluation model,which not only makes up for the insufficient of most existing research that the credit evaluation model neglected the default discrimination of new sample,but also avoids the tedious and time-consuming drawbacks of rebuilding evaluation model frequently.Using the 3 045 small businesses from a Chinese bank for empirical study,the empirical results show that the incremental credit evaluation model has high accuracy and thus is reasonable.