An intelligent bankruptcy prediction model using a multilayer perceptron

Abstract

High bankruptcy rates can lead to the collapse of economic systems. Therefore, having accurate and reliable models to predict firms in financial distress allows for proper management of the economic losses helping to prevent such crises. Since the 1930s, more than 500 studies have been published in the field of bankruptcy prediction models. In this paper, we firstly give a comprehensive literature review on the topic of statistical and intelligent models to predict firms failure. Then, we closely examine the discriminatory power of a Multilayer Perceptron (MLP) in the context of bankruptcy prediction. For this purpose, we consider different setups of optimization algorithms, activation functions, number of neurons, and number of layers. To find the parameter setup that achieves the best results, we use various evaluation metrics such as average accuracy, specificity, sensitivity, and precision. The case study is based on a data set of Taiwanese firms and includes comprehensive comparative analysis. The proposed MLPs show superior performance, and we critically examine the differences between the methodologies to explain the discrepancies in the results.

Keywords

• Artificial Intelligence (AI)
• Artificial Neural Networks (ANN)
• Business failure
• Data Science
• Deep Learning
• Machine Learning