Neural Network VS Genetic and Particle Swarm Optimization Algorithms in Bankruptcy

Articles in Press

Document Type : Original Article

Author

• Alireza Azarberahman

Assistant professor, Shandiz Institute of Higher Education, Mashhad, Iran

10.22067/ijaaf.2024.87788.1456

Abstract

Purpose: The goal of the study is to estimate an artificial neural network (ANN) model for bankruptcy prediction and optimize processes using the Particle Swarm (PSO) and Genetic (GA) algorithms.

Design/methodology/approach: 21 variables that were related to the likelihood of bankruptcy were chosen for the study. Neural networks (NNs) choose the optimal network with the least amount of error in training and evaluating patterns in the second phase. The neural network's weights and biases were optimized in the final stage by combining GA and PSO with the neural network.

Findings and contribution: The results showed that the initial pattern's ability to be explained has risen with the use of GA and PSO. The superiority of the models over linear regression is demonstrated by the evaluation of ANN performance. Finally, four variables—current ratio, sales to current assets ratio, economic value added, and gross profit margin ratio—that may reliably predict bankruptcy were found using the hybrid approach of ANNs-PSO and ANNs-GA. The evidence brings out the effectiveness of the metaheuristic algorithms as compared to linear ones on predicting bankruptcy. This goes on to further highlight the new breed of computational tools to available as techno-savvy financial analysts as well as investors.

Keywords

• Bankruptcy
• Artificial neural networks
• Genetic algorithm
• Particle Swarm algorithm

Main Subjects

• Finance

---