A dual-populations artificial bee colony algorithm

The artificial bee colony exists premature and convergent stagnation phenomenons when dealing with high dimension and complex problem. An effective dual-populations artificial bee colony (DABC) algorithm is presented to solve numerical optimization problem in this paper. This algorithm consists of two populations which uses different optimization strategy respectively. In one population that implements population global exploration, the employed bees and the onlookers adopt a combination of two strategies, i.e., artificial bee colony (ABC) operator and differential evolution (DE/rand/1) operators, and the scouts employ random search, which is good at population diversity. While in another population that aims to achieve solutions local exploitation, the employed bees and the onlookers use DE/best/2 differential operators, and the scouts employ adaptive search around the optimum to update nectar position, which is beneficial to quicken convergent speed. Two populations evolve independently with different optimization strategies, and the optimal nectars information of two populations are combined every certain generations to strive for a well balance between the global exploration and the local exploitation. In additional, during the onlookers selection nectar sources, non-linear ranking selection strategy is employed to alleviated premature problem that caused by super nectar in the colony. The experimental results of benchmark functions show that the DABC algorithm is superior to several other methods for the convergence speed and solution precision, which indicates this new algorithm is an effective approach for solving global optimization problems.