Exploring the Effects of Attraction and Repulsion Parameters on the Bacterial Foraging Algorithm through Benchmark Functions

Abstract

Metaheuristics are essential when working with complex problems from different fields. However, a suitable tuning scheme for these parameters is necessary to facilitate the search for potential solutions. This tuning is a challenging task. This work aims to develop a tuning method for the BFOA algorithm regarding attraction and repulsion values. In some cases, the parameter values are taken from previous works, while in other cases, the parametrization scheme comes from an automated or dynamic process. This work explores the Bacterial Foraging Algorithm (BFOA) within its parameters related to attraction and repulsion among bacteria, using 18 well-known benchmark functions from the literature. For this purpose, multiple BFOA executions were made, and averages were calculated for each test with repetitions for 24k BFOA executions. The interest variables for contrasting performance were the number of evaluated functions (NFE), the required time for the execution (time), and the associated cost to the achieved solution by BFOA (cost). Results: BFOA produced a different performance corresponding to each benchmark function. From this, four tuning schemes are proposed and validated by repetition, also contrasted by t-test. The conclusions show that the BFOA algorithm is susceptible to tuning, and the attraction and repulsion parameters must be according to the optimization problem. In terms of execution time, scheme III showed remarkable results. Regarding the obtained solution cost, scheme II outperformed the other three schemes.