Research On Early Warning Of Corporate Debt Default Based On RLasso-logistic Regression Model

With economic development and financial opening up,bond defaults have become the focus of attention in2020.In the bond market,general corporate bonds have more defaults.In particular,bond defaults in 2019 and 2020 have increased more than in previous years and have a greater impact on society.Therefore,Establishing a bond credit risk model suitable for my country's national conditions and bond market conditions,and studying the current status of bond defaults,is of great significance to bond entities,bond buyers,regulatory agencies,and governments.This paper uses 40 corporate bonds that defaulted in 2019 and 2020 as the research sample,and randomly selects 160 corresponding non-default bonds,with a total data volume of 200.At the same time,per share indicators,growth capacity,cash flow,profitability,capital structure,solvency,operating capacity and macro factors are included in the alternative indicators.Construct the r Lasso-Logistic model to explore the applicability of this model to the early warning of corporate debt default.And also used the Logistic model and the Lasso-Logistic model to conduct empirical research on the same data at the same time.Finally,the three models are compared,and the main conclusions are as follows:(1)In the case of multiple factors,especially the increasingly complex corporate bond default situation,there are more and more factors that can be taken into consideration.The simple logistic model cannot meet the research needs of bond default.(2)At present,there are few data in the bond market.The Lasso-Logistic model requires a lot of data,and it cannot exert its strong learning ability.It is not suitable for the research on early warning of corporate debt default and still needs improvement.(3)The r Lasso-Logistic model's requirement for the quantitative relationship between the sample size and the candidate factors,that is log < 9),is more in line with the current data volume of the corporate bond market;the model has good sparseness,which makes the judgment of the result more accurate;incorporating macro indicators into the built model makes the model more comprehensive.Therefore,the r Lasso-Logistic model is the best among the three models for the early warning of corporate debt default in China.