Bankruptcy prediction for Russian companies: Application of combined classifiers

• We use a dataset of more than 3,500 Russian manufacturing companies.
• We use different combined classifiers (LR, ANN, MDA, CRT) for bankruptcy prediction.
• Classical models of bankruptcy demonstrate low accuracy of bankruptcy prediction.
• Classifiers combined by AdaBoost approach show highest classification accuracy.
• Only one indicator stipulated by Russian legislation is effective for prediction.

The problem of bankruptcy forecasting is one of the most actively studied nowadays, posing the task of building effective classifiers as well as the task of dealing with dataset imbalance. In this paper, we apply different combinations of modern learning algorithms (MDA, LR, CRT, and ANNs) in order to try to identify the most effective approach to bankruptcy prediction for Russian manufacturing companies. Simultaneously, we try to find out whether the financial indicators stipulated by Russian legislation provide an effective set of indicators for bankruptcy prediction.