CORPORATE BANKRUPTCY PREDICTION USING MACHINE LEARNING TECHNIQUES

Abstract

Estimating the risk of corporate bankruptcies is of large importance to creditors and in- vestors. For this reason bankruptcy prediction constitutes an important area of research. In recent years artificial intelligence and machine learning methods have achieved promising results in corporate bankruptcy prediction settings. Therefore, in this study, three machine learning algorithms, namely random forest, gradient boosting and an artificial neural net- work were used to predict corporate bankruptcies. Polish companies between 2000 and 2013 were studied and the predictions were based on 64 different financial ratios. The obtained results are in line with previously published findings. It is shown that a very good predictive performance can be achieved with the machine learning models. The reason for the impressive predictive performance is analysed and it is found that the missing values in the data set play an important role. It is observed that prediction models with surprisingly good performance could be achieved from only information about the missing values of the data and with the financial information excluded.