Corporate Failure Prediction Using Machine Learning Models: Evidence from Nigerian Listed Companies

This study develops and validates corporate failure prediction models using machine learning techniques applied to financial data from companies listed on the Nigerian Exchange Group. Traditional statistical models such as Altman's Z-score and Ohlson's O-score were developed in mature market contexts and may not adequately capture the dynamics of financial distress in Nigerian emerging market conditions. Machine learning models, including random forest, support vector machine, and gradient boosting classifiers, offer potentially superior predictive accuracy through their capacity to detect complex non-linear relationships among financial variables. This study uses financial ratios derived from five years of pre-failure and control firm annual report data from 30 distressed and 30 non-distressed listed companies for the period 2015 to 2023. Features include liquidity, profitability, leverage, efficiency, and market-based indicators. Model performance is evaluated using accuracy, precision, recall, F1 score, and area under the ROC curve. Results indicate that the gradient boosting classifier achieves the highest predictive accuracy at 88 percent, significantly outperforming the traditional Altman Z-score at 71 percent in the Nigerian sample. The most predictive features were current ratio, interest coverage ratio, and earnings volatility. The study concludes that machine learning models offer a materially superior corporate failure prediction capability for Nigerian listed companies. It recommends that the Securities and Exchange Commission develop a machine learning-based early warning system to identify at-risk listed entities and trigger pre-emptive regulatory engagement.

Keywords: corporate failure prediction, machine learning, Nigerian Exchange Group, gradient boosting, financial distress.