Comparative Study of Reduced Parameter Versions of the Bees Algorithm for Traveling Salesman Problem

Metaheuristics have shown dominance over exact methods with their capability to find near-optimal solutions to complex problems in a shorter time. Among these metaheuristics, the Bees Algorithm (BA) has proven its performance in various applications. However, fine-tuning the parameters of the BA is challenging due to its numerous parameters. There have been few studies aiming to reduce the number of parameters while maintaining or improving performance, such as the ternary BA, two-parameter BA, and Fibonacci BA. This paper reviews these variants for combinatorial problems using 13 datasets from the Travelling Salesman Problem TSPLIB. The results were compared using an independent t-test in conjunction with descriptive statistics. The findings show that the Fibonacci BA outperforms other variants, and potential suggestions for improvements in the future were proposed.