Determinative Brain Storm Optimization

Abstract

Brain Storm Optimization (BSO) is a swarm intelligence optimization algorithm, based on the human brainstorming process. The ideas of a brainstorming process comprise the solutions of the algorithm, which iteratively applies solution grouping, generation and selection operators. Several modifications of BSO have been proposed to enhance its performance. In this paper, we propose a novel modification enabling faster convergence of BSO to optimal solutions, without requiring setting an upper bound of algorithm iterations. It considers a brainstorming scenario where participating groups with similar ideas recognize that their ideas are similar, and together, collaborate for the determination of a better solution. The proposed modification, called Determinative BSO (DBSO), implements this scenario by applying a cluster merging strategy for merging groups of similar solutions, while following elitist selection. Experimental results using eleven benchmark functions show that the proposed modified BSO performs better than both the original and a state-of-the-art algorithm.