Artificial Bee Colony Algorithm with Population Reduction Strategy

Population size (NP) has always been a very important parameter in artificial bee colony (ABC) algorithm. It is found that large population size facilitates exploration and ensures population diversity, while small population size facilitates exploitation and saves computing resource. Therefore, dynamically reducing the population size during the search is helpful for the optimization process of the algorithm. This paper draws lessons from the population reduction strategies of other algorithms including linear reduction, half reduction and their combinations with K-means clustering, and applies them to the artificial bee colony algorithm. The experimental results on test set including continuous unimodal function, noise function and multimodal function about solution accuracy, stability and convergence speed proved the different effectiveness of the population reduction strategies for ABC algorithm. It is further found that the population reduction strategy with K-means clustering is better than that without K-means clustering.