Comparative analysis of machine learning models for bond default forecasting based on financial data of Chinese listed companies

Abstract

At a time when bond defaults become frequent and market confidence is undermined, the subject of how to accurately predict bond defaults merits investigation. This paper combined popular machine learning methods to predict bond defaults by selecting financial data of 23 listed companies in China that defaulted on credit bonds from 2013 to 2022 and 230 financial data of listed companies that did not default on credit bonds during the same period, using logistic regression, random forest, support vector machine (SVM), and K Nearest Neighbors (KNN) to estimate the probability of bond defaults by listed companies and compare the predictive performance of these methods. The financial data are found to be quite good at predicting bond defaults of listed companies, and the SVM model performs the best.