Using Machine Learning for Predicting Business Failure: A Systematic Literature Review

One of the pioneer topics in the finance industry is the prediction of bankruptcy using machine learning (ML) methods. Bankruptcy prediction is an essential indicator for various organizations, including banks, corporations, countries, and many more. These financial organizations identify specific strategies when bankruptcy prediction is determined. Therefore, financial institutions must ensure the accuracy of the models used. The ML methods have proven the relevance and robustness of the models applied in predicting the bankruptcy of organizations within different regions of the world. This Systematic Literature Review (SLR) covers 38 papers on bankruptcy prediction using ML. This study explores ML advancements in predicting business failure in Eastern and Central Europe and Gulf Cooperation Countries (GCC) regions. In addition, this paper aims to examine the technical issues and pitfalls concerning predicting models applied to bankruptcy. The third purpose is to validate the superior power of ML in achieving forecasting goals on the topic. These studies have resulted in a spectacular performance of models produced by ML methods in the context of bankruptcy. However, some issues related to data and model setting affect the quality and performance of the models, such as selecting variables, imbalanced data, and outliers. In addition, the findings demonstrate that it's essential to keep in mind and consider that the characteristics of some nations have an impact on how the study field covered in this article evolves. Moreover, imitating other models without considering the nations' specificities may not lead to the best findings.