Parameter Tuning for Combinatorial Bees Algorithm in Travelling Salesman Problems

Abstract

Bees Algorithm is one of the most used nature-inspired algorithms. There are five parameters applied in the basic combinatorial version of Bees Algorithm: number of scout bees, number of elite bees, number of best bees, number of elite sites, and number of best sites. Parameter tuning is one of the critical and time-consuming steps in metaheuristic algorithms. This research is the first parameter tuning study for Combinatorial Bees Algorithm (BA) for solving the Travelling Salesman Problem (TSP). The experiments are designed using Fractional Factorial Design, and four steps, including parameter setting and statistical analysing, are carried out. The TSP problem’s goal is to minimise the total path and find the lower number of the best cost. Comprehensive experiments have been done using varying TSPLIB datasets between 51 and 575 cities to minimise the total path and find the lower number of the best cost. Statistical results show that the best combinatorial BA parameters are the balanced scenario of local and global search.