An Empirical Study On The Influencing Factors Of Corporate Bond Credit Spread In China Based On Machine Learning

Since the "11 Super Day Bond" incident in 2014,which led to the breaking of the rigid bond convention,the default rate of my country’s credit bond market has been increasing year by year,and corporate bonds,as one of the three basic credit bonds,occupy an important position in the credit bond market.Therefore,it is of great significance to study the credit spread of corporate bonds.This paper first selects the corporate bonds issued in my country’s primary bond market from January 1,2016 to February 28,2021 as a sample,and combines the existing literature on credit spreads and related economic theories to screen out from the macro,micro and bond levels.The 24 more important influencing factors of corporate bond credit spread are used as explanatory variables of the empirical model.Secondly,multiple regression,random forest,XGBoost and Light GBM algorithms are used to fit and predict corporate bond credit spreads,and compare the prediction performance of different models.Finally,we use interpretable machine learning techniques such as feature importance,PDP,ALE,and SHAP value to perform interpretable analysis on the influencing factors of credit spreads.The research results show that: 1.Machine learning models can be used to predict and feature selection of corporate bond credit spreads well,and they all have very good prediction performance.Among them,the prediction performance of XGBoost model based on feature selection is all the predictions in this paper.The best of the models.2.The corporate bond credit spread is mainly affected by eight characteristic variables,including risk-free interest rate,issuance period,long-term and short-term treasury bond interest rate,stock market index,money supply,economic growth level,inflation level and exchange rate,and the risk-free interest rate is the key factor.Among them,the issuance period belongs to the bond factor,and the other seven characteristic variables are macro factors.3.Through interpretable machine learning technologies such as SHAP,it is possible to conduct a more detailed analysis of the relationship between corporate bond credit spreads and their influencing factors,so as to make more accurate judgments on changes in corporate bond credit spreads.