Financial Distress Modelling and Prediction in Emerging Asia: A Machine Learning Analysis of India, China and Korea

Abstract

This study models corporate financial distress for three large emerging Asian markets, India, China and Korea, over 2011 to 2024 using firm data from the LSEG database. Five classifiers, a binomial logit and four machine learning models, are estimated on a lean predictor set chosen through an information gain procedure across six ratio categories. Predictors selected at the pooled level are first applied to each country, and the exercise is then repeated with predictors chosen separately for each market. Random Forest and Gradient Boosting lead in every market. India records the highest accuracy and gains most from country-specific predictors, where Random Forest rises to 92.2 per cent. China is stable and Korea remains the hardest to classify, a pattern consistent with its concentrated ownership structure.