A novel artificial bee colony algorithm based on the cosine similarity

Abstract Artificial bee colony (ABC) is a very popular and powerful optimization tool. However, there still exists an insufficiency of slow convergence in ABC. To further improve the convergence rate of ABC, a novel ABC (CosABC for short) is proposed based on the cosine similarity, which is employed to choose a better neighbor individual. Under the guidance of the chosen neighbor individual, a new solution search equation is introduced to reduce the weakness of undirected search of ABC. Furthermore, in the employed bees phase, a solution search equation with the guidance of global best individual is also integrated, and the frequency of parameters perturbation is also employed to further increase the information share between different individuals. In the onlooker bees phase, ABC/rand/1/ is used to enhance the exploitation ability, yet an opposition-based learning technique is also used to balance the exploitation of ABC/rand/1. All these modifications together with ABC form the proposed CosABC algorithm. To demonstrate the effectiveness of CosABC, a comprehensive experimental research is conducted on a test suite composed of twenty-four benchmark functions. What is more, it is further compared with a few state-of-the-art algorithms to validate the superiority of CosABC. The related comparison results show that CosABC is effective and competitive.