Bankruptcy Prediction through Ensemble Machine Learning-Based Risk Assessment

Abstract

Bankruptcy is one of the biggest threats to a company's reputation, occurring when it is unable to pay back outstanding debts to banks, lenders, and suppliers. Predicting bankruptcy accurately and promptly allows companies to take remedial action in advance and avoid it. To achieve this goal, current research is investigating a combination of techniques that can accurately predict bankruptcy. The proposed method employs various ensemble techniques to combine the best methods for improved accuracy. The proposed ensemble models have been compared using both the original imbalanced dataset and the balanced dataset created by oversampling. The proposed method of using the balanced dataset on the ensemble models outperformed the models using the original dataset in accuracy and other performance metrics. Balance Bagging achieved the highest accuracy at 98.77%, followed by Random Forest at 98.68%, and AdaBoost at 96.9%. These results are a significant achievement compared to state-of-the-art techniques.