| As the basic force to promote market economic development,small enterprises have made great contributions to economic growth,employment increase and social stability.However,due to the weak foundation of small enterprises,with the characteristics of small scale,financial opacity,non-standard operation and relatively single financing channels,small enterprises are facing the difficulties in financing,which seriously impedes the development of small enterprises.In view of the current situation of small enterprises,it is very necessary to construct a set of scientific and reasonable credit rating methods suitable for their own characteristics,which can largely solve the problem of information asymmetry between banks and enterprises,so as to alleviate the financing difficulties of small enterprises.Most of the existing credit ratings take large enterprises as the research object,although some small enterprises are also considered,compared with large enterprises,the credit rating method of small enterprises is not mature and cannot accurately reflect the credit status of small enterprises.Therefore,the credit rating method of small enterprises should be further discussed to provide financial support for the healthy development of small enterprises.Main works of this paper:First,a credit rating indicator system with both sensitivity and discrimination is constructed.The non-parametric Bayesian model is used to conduct the first stage discrimination of small enterprises with default and non-default,and the credit indicators that can significantly distinguish small enterprises with default and non-default are screened.Based on the small enterprise credit rating indicator system screened in the first stage,non-parametric clustering is used to cluster the defaulting small enterprises into small enterprises with high loss given default and low loss given default,and then the non-parametric Bayesian model is used to conduct the second stage discrimination on small enterprises with high default and low default,and the credit indicators that can significantly distinguish high default and low default of small enterprises are screened.The empirical results show that the credit rating indicator system of small enterprises this paper finally constructed consists of 24 indicators,such as the quick ratio,cash ratio and gross margin.Compared with the credit rating indicator system constructed by two-stage parameter Bayesian discriminant model and two-stage Logistic regression model,the results show that the two-stage non-parametric Bayesian discriminant model is the best,the two-stage Logistic regression model is the second,and the two-stage parameter Bayesian discriminant model is the worst.Second,the portfolio credit scoring model with default discrimination ability is constructed.On the basis of the indicator system finally constructed in the previous chapter,three-classification non-parametric Bayesian discrimination is selected as the representative method of mathematical statistics model,ordered three-classification Logistic regression as the representative method of econometric model,and three-classification support vector machine as the representative method of artificial intelligence.Through the three-classification non-parametric Bayesian discrimination to measure the weight of indicators,it reflects the weight calculation thinking that the greater the impact of indicators on the discrimination accuracy,the more important the indicators are.Through the ordered three-classification Logistic regression to measure the weight of indicators,it reflects the weight calculation thinking that the greater the impact of indicators on default status,the more important the indicators are.And through three-classification support vector machine to measure the weight of indicators,it reflects the weight calculation thinking that the greater of non-default,low default and high default distance is,the more important the indicators are.Then,by constructing an optimization model that minimizes the sum of squares of deviation between the portfolio credit score results and the single method credit score results to aggregate the similar characteristics of these three methods.Finally,the credit score of each small enterprise is obtained by linear weighting method.The empirical results show that the investment return rate has the largest weight,and the legal representative loan default record has the smallest weight.The credit scores of small enterprises are mostly between 50 and 60,and the credit score results have a downward bias.Third,the credit rating model of optimal discriminant ability is constructed.Based on the small enterprise credit scores obtained in the previous chapter,the maximum ratio of the sum of the dispersion of credit scores between different credit grades and the sum of the dispersion of credit scores within the same credit grade as the objective function,and the loss given default of the following credit grade is strictly gerater than the loss given default of the previous credit grade as the inequality constraint,a nonlinear credit rating optimal partition model is constructed.And the approximate solution of the model is solved by a recursive algorithm with strong reproducability and clear structure.The empirical results show that when the objective function threshold critical value is 10000 and the maximum value of objective function value is 9932.76,the credit score intervals of the nine credit grades from AAA to C are,respectively,[53.61,100.00],[49.69,53.61),[44.81,49.69),[37.87,44.81),[16.91 37.87),[14.22,16.91),[8.69,14.22),[5.86,8.69),[0,5.86),the corresponding loss given default are,respectively,0.19%,8.21%,10.49%,16.05%,18.39%,52.32%,69.04%,94.55%,95.74%.The comparative analysis based on the customer number distribution and K-means clustering show that the credit rating model constructed in this paper is the best,while the credit rating model based on the customer number distribution and K-means clustering are poor.The innovation of this paper:First,By combining whether there is default with different default losses,a new idea of credit rating indicator screening based on two-stage Bayesian discrimination is proposed,and a credit rating indicator system with stronger sensitivity and greater discrimination is constructed,which can not only distinguish the two different credit characteristics of default and non-default,high default and low default,but also ensure that the screened indicators have more intuitive and in-depth discrimination of default samples.Moreover,it makes up for the deficiency that most existing research ignore the influence of credit characteristics of different nature,such as total default and partial default,on indicator screening.Second,by constructing an optimization model that minimizes the sum of squares of deviation between the portfolio credit score results and the single method credit score results to aggregate the similar characteristics of three representative methods of three-classification nonparametric bayesian discrimination,ordered three-classification logistic regression and three-classification support vector machine.It solves the problem that credit score results changing with each single method,integrates the advantages of each method and improves the credibility,reliability and consistency of credit score result.Third,based on the principle that the credit score gap of small enterprises within the same credit grade is the smallest and the credit score gap of small enterprises between different credit grades is the largest,the nonlinear optimization model of small enterprise credit rating is constructed,which not only satisfies the principle that the credit grade matches the loss given default,but also satisfies the principle that the credit group of small enterprises matches the credit grade.It also ensures that the small enterprises with small credit score gap are of the same credit grade,while the small enterprises with large credit score gap are of different credit grades,which overcomes the disadvantages of the existing research that only consider the small enterprises with large credit score gap and ignore the small enterprises with small credit score gap. |
PDF Full Text Request