Abstract:
This study predicts bankruptcy among Indian firms using artificial intelligence-machine learning (AI-ML) methods, demonstrating their superior performance over traditional statistical models. Addressing class imbalance through oversampling techniques such as the synthetic minority oversampling technique (SMOTE), the paper achieves higher accuracy rates with AI-ML models, such as random forest, neural networks, and gradient boosting. By leveraging the information value and weight of evidence of explanatory variables, the study designs early warning variables for bankruptcy risks, helping both internal and external stakeholders to monitor and mitigate these risks. The analytical framework thus extends the methodological application of AI-ML models and offers a management toolkit that practitioners can use to track and address bankruptcy risks effectively. Furthermore, the study finds that AI-ML models improve prediction accuracy, especially for listed firms, because of better information content in their financial statements.
