Predicting Financial Distress of Chinese Listed Firms Using Machine Learning Models: A Comparative Study
TL;DR
The results demonstrate that ensemble methods, particularly Gradient Boosting Machine, achieve superior predictive accuracy compared to traditional statistical approaches and contribute to both academic literature and practical applications in credit risk assessment within the Chinese regulatory context.
Abstract
Financial distress prediction remains a critical concern for investors, regulators, and stakeholders in China's rapidly evolving capital markets. This study presents a comprehensive comparative analysis of machine learning models for predicting financial distress among Chinese listed firms. Using a dataset of 2,847 Chinese A-share listed companies from 2015 to 2023, this study evaluates the predictive performance of Logistic Regression, Random Forest, Support Vector Machine (SVM), and Gradient Boosting Machine (GBM). The results demonstrate that ensemble methods, particularly Gradient Boosting Machine, achieve superior predictive accuracy (94.2%) compared to traditional statistical approaches (Logistic Regression: 86.7%). Feature importance analysis reveals that profitability ratios, leverage metrics, and liquidity indicators constitute the most significant predictors of financial distress. The findings contribute to both academic literature and practical applications in credit risk assessment within the Chinese regulatory context.