MACHINE LEARNING IN BANKRUPTCY PREDICTION – A REVIEW
MACHINE LEARNING IN BANKRUPTCY PREDICTION – A REVIEW

Summary/Abstract: There is an increasing interest in machine learning for bankruptcy prediction with more and more researchers contributing to the literature. Although there is a considerable amount of research, the domain does not seem to be aligned and there is still a lot of indecisiveness in terms of what is the best method to be used and on which data. Using Web of Science, Scopus and ScienceDirect databases, a systematic review of 32 texts published between 2016 and 2020 was conducted. This review shows a summary of those papers based on 9 criteria. The criteria identified include source of data, number and type of variables, models used, industry type, and timeline of dataset, sample size, aim and result as well as accuracy of the best performing model used. Overall, it has found that no model performs best on any type of data and that the domain is still away from having a conclusion about what works best and where. This paper contributes towards updating academics and practitioners with the current state of the domain, tools used for bankruptcy prediction lately and their performance.

• Details
• Contents