A multi-strategy fusion artificial bee colony algorithm with small population

Abstract Although artificial bee colony (ABC) algorithm is more and more popular in solving complex problems, slow convergence rate limits its wide application. ABC with small population can use the limited function evaluation times more efficiently since it can avoid unnecessary searches. However, ABC with small population cannot ensure population diversity, and when the algorithm is weak or unstable, it may fall into local optimum easily. So based on the latest research, we are motivated to propose a stabler and more efficient algorithm design to improve the search ability of ABC with small population by the fusion of multiple search strategies, which used together for the employed bees and the onlooker bees. Firstly we select and design multiple strategies with different search abilities of exploration and exploitation. Secondly, we propose an evolution ratio, which is an indicator to fully reflect the adaptability of the search strategy. Thirdly, we design different fusion methods according to the characteristics of the strategies, in which the search strategy with high exploration is maintained at a certain frequency throughout the whole search process of the employed bees, and the selections of the other two search strategies are adjusted according to evolution ratio adaptively in the employed bee phase and the onlooker bee phase. In the end, a novel algorithm called MFABC is proposed, which can realize efficiently multi-strategy cooperative search according to the requirements of different problems and different search stages. Experimental results on a set of benchmark functions have shown the accuracy, stability, efficiency and convergence rate of MFABC.