The Application of Ensemble Methods in Forecasting Bankruptcy

In practice, one chosen method is generally used to solve classification tasks. Although the most modern procedures yield excellent accuracy rates, international research findings show that a concurrent (ensemble) application of methods with weaker classification performance achieves comparable rates of high accuracy. This article's main objective is to compare the predictive power of the two ensemble methods (Adaboost and Bagging) most commonly used in bankruptcy prediction, using a sample consisting of 976 Hungarian corporations. The article's other objective is to compare the accuracy rates of bankruptcy models built on the deviations in specific financial ratios from industry averages to those of models built on financial ratios and variables factoring in their dynamics