Company Bankruptcy Prediction with SMOTE

Abstract

Abstract: Effective bankruptcy forecasting is essential for financial firms to make sound loan decisions. In general, the two most essential aspects influencing prediction performance are input variables (or features) such as financial ratios and prediction approaches such as statistical and machine learning techniques. While numerous relevant publications have presented innovative prediction algorithms, relatively few have examined the discriminating potential of bankruptcy prediction variables. In addition to financial ratios (FRs), corporate governance indicators (CGIs) have been identified as an important form of input variable in the literature. The prediction performance produced by merging CGIs with FRs, however, has not been thoroughly investigated. Only a subset of CGIs and FRs were employed in related investigations, and the characteristics used may vary from study to study. As a result, the goal of this work is to evaluate the prediction performance obtained by combining seven distinct FR categories and five different CGI categories. Based on a real-world dataset from Taiwan, the experimental results reveal that the FR categories of solvency and profitability, as well as the CGI categories of board structure and ownership structure, are the most critical variables in bankruptcy prediction. The best prediction model performance is determined by combining prediction accuracy, Type I/II errors, ROC curve, and misclassification cost. However, these findings may not be applicable in certain countries, such as China, where the definition of distressed enterprises is vague and the features of corporate governance measures are not explicit.