Systematic review of bankruptcy prediction models

The current research aims to systematically examine bankruptcy prediction models with the goal of developing a model that serves as a guide for selecting the most suitable tools. These tools should ideally align with the existing data and quality criteria of bankruptcy prediction models.

To conduct this research, a systematic search was performed on the Web of Science database using keywords such as Bankruptcy, Default, Distress, Failure, Forecasting, Predicting, Prediction, and Insolvency, spanning the years 2015 to 2023. Based on defined inclusion and exclusion criteria, this search yielded 1000 articles, out of which 49 were ultimately selected and analyzed. The findings from these articles were then summarized in tables. Subsequently, major bankruptcy prediction models were compared based on nine key criteria, and final conclusions were drawn.

Artificial neural networks and support vector machines were found to have the highest accuracy, while multiple personality analysis showed the lowest accuracy. Additionally, artificial neural networks, multiple personality analysis, decision trees, and logistic regression require a large training sample to logically identify and precisely classify patterns. However, case-based reasoning, rough sets, and support vector machines can work with smaller sample sizes.

Originality/ Value:

 The outcomes of this research contribute to a comprehensive understanding of the characteristics of tools used in developing bankruptcy prediction models and the shortcomings associated with them.