An effective refined artificial bee colony algorithm for numerical optimisation

Various complex problems have recently encouraged research and development of different bio-inspired optimisation algorithms, a well-known instance being the artificial bee colony (ABC) algorithm, both due to its simplicity and performance. Building upon the basic algorithm enabled further gains in performance but brought alongside it some specific costs and problems. The improved variants available in the literature often introduce additional user-defined parameters and sometimes completely infringe the algorithm structure. Focusing the search process on exploitation has proven to be a good first step of improvement in most cases, but analysing the effects of this modification on a limited set of standard benchmark functions could lead to a skewed perspective. This paper proposes a novel algorithm based on ABC that keeps the original structure intact, introduces a new solution update equation and an extended scout bee phase focusing the search on more prominent solutions without introducing new control parameters. Based on the conducted experimental analysis, it is able to outperform various competitive algorithms on a large test bed of benchmark functions and several real-world problems. The effects of the particular proposed modifications are also analysed and attention is given to two variants of the standard algorithm. (C) 2019 Elsevier Inc. All rights reserved.