Improved artificial bee colony algorithm with dynamic population composition for optimization problems

The artificial bee colony (ABC) algorithm is an effective swarm-based meta-heuristic algorithm for optimization problems. Nevertheless, slow convergence speed has affected its competitiveness. In order to improve its performance, an improved ABC with dynamic composition (ABCDC) is proposed in this paper. Since the original ABC and its most variants use constant ratio between employed bees and onlooker bees, which causes that the number of onlooker bees is insufficient to exploit the searching space in limited time. Therefore, we propose a mechanism to adjust the number of employed bees and onlooker bees in order to find the global optimum more effectively. Moreover, Symmetric Latin Hypercube Design is utilized to enhance the diversity of initial population. Besides, two differential search equations with self-adaptive parameters are used in the employed bee phase and onlooker bee phase. Finally, to evaluate the performance of ABCDC, comparisons with four state-of-the-art ABC variations and the original one have been done on 22 benchmark problems with different dimensions. And four meta-heuristic algorithms were also involved to fully evaluate the effectiveness of ABCDC. The experimental results demonstrate that ABCDC is better than the competitors in terms of its solution quality and convergence speed.

Automated summary

The improved ABC algorithm proposed in this paper adjusts the number of employed bees and onlooker bees, enhances the diversity of initial population, and utilizes self-adaptive parameters in differential search equations, which leads to better solution quality and convergence speed compared to competitors.