A Multi-Objective Artificial Bee Colony Algorithm Combined with a Local Search Method

Abstract

This paper proposes a new multi-objective artificial bee colony (ABC) algorithm called MOABCLS by combining ABC with a polynomial mutation based local search method. In this algorithm, an external archive is used to store the non-dominated solutions found so far which are maintained by the crowding distance method. A global best food source gbest is selected and used to produce new food sources in both employed and onlooker bee phases. The aim of adopting a local search is to keep good balance between exploration and exploitation. The MOABCLS is able to deal with both unconstrained and constrained problems, and it is evaluated on test functions (with up to five objectives) taken from the CEC09 competition. The performance of MOABCLS is compared with that of eight state-of-the-art multi-objective algorithms with respect to IGD metric. It is shown by the Wilcoxon test results that MOABCLS performs competitively or even better than the peer algorithms. Further experimental results clearly demonstrate MOABCLS’s ability of finding a set of well converged and appropriately distributed non-dominated solutions, and the performance promotion by introducing the local search method.