Research On Multi-class Imbalanced Corporate Bond Default Risk Prediction Based On The Adaboost Ensemble Model

Since the first material default in 2014,there has been an increase in bond defaults,which have attracted widespread attention in society and increased investors’ vigilance.Because of the asymmetry of information,investors do not know the real corporate bond default risk,which is not conducive to making correct investment decisions,so corporate bond default risk prediction can effectively help investors to identify risks in a timely manner,and it is also important for bond market regulators and issuers.But most of the current corporate bond default risk prediction studies are two-class predictions,and most of them use a single classifier model.In this paper,a new method for predicting multi-class imbalanced corporate bond risk is proposed,which combines the synthetic minority over-sampling technique(SMOTE)and the Adaboost ensemble method to solve the class-imbalanced problem of samples,and uses oneversus-one(OVO)decomposition method and one-versus-all(OVA)decomposition method to transform multi-class classification problems into two-class classification problems.We categorize corporate bond default risk into three classes based on the credit rating of the bond issuer: very low default risk,relatively low default risk and high default risk,which is more scientific than the traditional two-class bond default risk,and carry out empirical experiments by respectively using DT,SVM,Logit and MDA as basic classifiers.Empirical results show that the OVO-SA-DT model has the optimal and stable overall prediction performance in all the integrated models,which is more suitable for multi-class imbalanced corporate bond default risk prediction.In addition,the empirical results of the integrated model and the single classifier model show that the overall performance,the bond recognition ability with very low default risk and the bond recognition ability with low default risk of the OVO-SA-DT model are all much better than the OVO-S-DT model,and the OVA-SA-DT model has better overall performance than the OVA-S-DT model,as well as the bond recognition ability with low default risk and the bond recognition ability with high default risk.Therefore,when DT is the basic classifier,it is important to integrate the Adaboost algorithm into the integrated model of multi-class imbalanced corporate bond default risk prediction.This paper’s research on the multi-class imbalanced corporate bond default risk prediction can not only help enterprises discover their own crises in advance,take timely countermeasures,and do a good job in advance prevention,but also provide bond market regulators with more effective risk warning method,reduce investors’ losses,and promote the stable development of the bond market.