Research On Credit Risk Of Listed Companies Based On Quantile Regression

With the research methods of fields of credit risk in China and abroad deepening,the sharp fluctuation of the domestic capital market once again aroused the attention of the commercial banks,investors and government regulators on the credit risk of listed companies.Therefore,based on the summary of research methods in China and abroad,the Binary Bayesian Quantile Regression model is applied to the field of credit risk identification and measurement,and respectively compare with Binary Logit regression model and KMV model.The model will selected some A shares listed companies from 2014 to 2016 which were all treated with ST or *ST special publication as low credit risk group.According to the industry and the market value of corporate defaults and other factors,choose the corresponding A shares listed companies as low credit risk group.The prediction time is 1 years.All involved in the regression model's data whether financial information data or capital information data were selected T-2(T for special treatment of publicity year)data points on time.On the basis of the above samples and indicators,we make an empirical analysis and the expain of resukts of Binary Bayesian Quantile Regression model,Binary Logit model and KMV model.A regression analysis of the regression results of the Binary Bayesian quantile regression model at different points is given,and the two sample nonparametric test is used to analyze the default distance of the KMV model.Finally,the introduction of a series of evaluation indicators and evaluation curves such as accuracy,recall,ROC-curve and PR-curve are placed.A comparative study was made between the two models and the model by using the Binary Bayesian quantile regression model and the Logit regression model and the KMV model.Through empirical analysis and comparative study indicate that: Binary Bayesian quantile regression model can effectively improve the Binary Logit model which is analyzed only for the relationship between the average risk and the financial indicators,We obtain the change tendency of each financial index under the condition of different credit risk,at the same time through comparative study three models,we can be concluded that the comprehensive level of Binary Bayesian Regression model is the highest.