Swarm Intelligence Algorithm with Guided Exploitations: A Case Study with Artificial Bee Colony Algorithm

During any meta-heuristic search, two opposite processes are found in action, namely the explorations and exploitations. Although they might seem to operate in opposite directions, they are actually counterparts, and synergy between them may improve the final outcome of the algorithm. This is especially true for complex, high dimensional problems, because the search algorithm has to avoid many local optima to find a good near optimum solution. There exist many swarm intelligence algorithms that report the necessity of a proper balance between explorations and exploitations. This paper presents a concrete example of a swarm intelligence algorithm, i. e. , the Artificial Bee Colony (ABC) algorithm that finds improvement by balancing between explorations and exploitations. In this paper, we have introduced ABC with Guided Exploitations (ABC-GE), a novel algorithm that improves over the basic ABC algorithm. ABC-GE augments each candidate solution with a control parameter that controls the proportion of explorative and exploitative perturbations and thus affects how new trial solutions are produced from the existing ones. This control parameter is automatically adjusted at the individual solution level, separately for each candidate solution xi, to adjust the proportions of explorations and exploitations around xi. ABC-GE is tested on a number of benchmark problems on continuous optimization and compared with the basic ABC algorithm. Results show that the performance of ABC-GE is overall better than the basic ABC algorithm.