Predicting Corporate Bankruptcy in Listed Companies Using Artificial Intelligence Algorithms and Financial Statement Data

Abstract

This study aimed to predict the bankruptcy risk of firms listed on the stock exchange using machine learning algorithms and financial statement data and to identify the most influential predictors of bankruptcy risk. This applied quantitative study employed a descriptive–analytical design. The statistical population consisted of companies listed on the stock exchange during the 2018–2024 period. Eligible firms were selected through a screening process based on data availability and consistency criteria. Financial data were extracted from audited financial statements, annual reports, and financial databases. Two machine learning algorithms, Random Forest (RF) and Balanced Random Forest (BRF), were used to predict bankruptcy risk. Model performance was evaluated using Area Under the Curve (AUC), Accuracy, Precision, Recall, and F1-score. Feature Importance analysis was conducted to determine the contribution of each predictor, while the Generalized Method of Moments (GMM) was employed to examine dynamic relationships among variables. The results indicated that the Random Forest model achieved an overall accuracy of 0.855 with an AUC of 0.453; however, it showed limited ability to identify bankrupt firms. In contrast, the Balanced Random Forest model demonstrated superior performance in detecting financially distressed companies despite a slightly lower overall accuracy (0.824). The BRF model yielded Precision, Recall, and F1-score values of 0.333, 0.211, and 0.258, respectively. Confusion matrix results further confirmed the greater sensitivity of the balanced model toward the minority class. Feature importance analysis revealed that institutional ownership, accounts turnover ratio, earnings-to-price ratio, and cash-to-sales ratio were among the most influential variables in predicting bankruptcy risk. The findings suggest that machine learning techniques, particularly the Balanced Random Forest algorithm, provide an effective framework for the early identification of firms facing bankruptcy risk. Integrating financial indicators with ownership structure and corporate governance variables can enhance predictive performance and support more informed decision-making by investors, managers, and financial analysts.