Predicting Corporate Financial Failure Using Sigmoidal Opposition-Based Arithmetic Optimization Algorithm

This paper concentrates on solving corporate financial failure prediction problems using a novel method. Corporate financial failure prediction is considered as a high complexity problem. It is hard to solve with traditional prediction algorithms. Notwithstanding, metaheuristics are aimed to solve these types of problems. Among them, is the Arithmetic Optimization Algorithm (AOA), which is one of the newest metaheuristics that is characterized by its easy integration, usability and strong computational ability. It is estimated to be one of the most used metaheuristics. In this paper, we propose an improved version of it called Sigmoidal Opposition-based Arithmetic Optimization Algorithm (SOAOA) in which the Opposition-based Learning is applied to improve the local searching capability and boost the intensification phase of the AOA. Whereas, the integration of the sigmoidal function enhances its diversification phase that results in better outcomes. The main purpose of this paper is to present our algorithm, which has proven to provide highly accurate results in predicting bankruptcy. In order to verify the latter, we have applied it to 50 well-known benchmarking functions to see how it deals with global optimization. Then we compared it with the most popular and exact Machine Learning algorithms such as Support Vector Machine (SVM) and Decision Trees (DT) to determine its accuracy in solving the formerly mentioned problem. SOAOA results are prominent in both global optimization and bankruptcy prediction tests. Based on the results, it has shown to be the best algorithm for solving the task evenly with DT.