A critical analysis of the bat algorithm

Abstract

This article presents an analysis of the bat algorithm (BA) based on elementary mathematical analysis and statistical comparisons of the first hitting time performance metric distributions obtained on a test set comprising five carefully selected objective functions. The findings show that the BA is not an original contribution to the metaheuristics literature and that it is not generally superior to the Particle Swarm Optimization algorithm when fair comparisons are made. It is also shown that some components of the BA can be either replaced by simpler alternatives or be removed entirely to increase performance. Finally, the results suggest that the best version of the BA is in fact a simple hybrid between Particle Swarm Optimization and Simulated Annealing. To encourage more transparency in metaheuristics research, the entirety of the MATLAB code used in this article is available in a GitHub repository for suggestions and/or corrections.