Evaluation Of Credit Risk Of Listed Companies In China Based On Sparse Principal Components-logistic Model

Our country's existing credit risk evaluation research has two main problems: First,the input variable information is redundant and redundant,which not only affects the evaluation effect of the credit risk evaluation model,but also increases the amount of calculation of the credit risk evaluation process.Second,the evaluation index system is not comprehensive enough.Existing research mainly considers financial indexes and ignores the role of non-financial indexes.Therefore,it is of great practical significance to study the credit risk evaluation model to improve the accuracy and stability of the model.In view of this,this paper uses a combination of sparse principal components and Logistic model to evaluate credit risk.First,this paper proposes a cr edit risk evaluation method based on the sparse principal component-Logistic model.The sparse principal component method is used as a dimensionality reduction tool.Through linear transformation,several indicators are converted into independent principal component factors,and most of the data features are concentrated into a few Each principal component factor;input the principal component factor into the Logistic model for credit risk evaluation,and improve the accuracy of the model.Secondly,to analyze the current status and causes of the credit risk of listed companies in China,it is proposed to include non-financial indicators in the existing evaluation index system for the problem that the existing credit risk evaluation index system is not comprehensive.Other indicators are included to make a more comprehensive evaluation of the listed company's credit risk.The empirical study mainly selected 164 ST companies among my country's Shanghai and Shenzhen A-share listed companies from 2016 to 2018 as credit default samples,and selected 492 non-ST listed companies as non-default samples using a sample matching ratio of 1:3.Combined with statistical test methods,25 financial indicators with significant differences in the two groups of samples were sc reened out,and three non-financial indicators including company size,shareholding ratio of the largest shareholder and equity pledge were added to construct a Logistic model based on sparse principal components Carry out credit risk evaluation and foreca st to verify the performance of Logistic regression model in terms of forecast accuracy and stability.The model in this paper has the following advantages: First,the inclusion of non-financial indicators makes the evaluation system model more comprehensi ve;Second,the principal component method reduces redundant information between multiple financial indicators and reduces the negative effects of information redundancy;Third,reduces the risk evaluation process The amount of calculation is convenient for the interpretation of economic significance.Empirical research shows that the Logistic model based on sparse principal components has good stability and high accuracy for the prediction of credit risk of listed companies in my country,and the prediction accuracy can reach more than 90%.Finally,according to the empirical results,some corresponding policy suggestions are put forward.