An Improved Artificial Bee Colony Algorithm with Elite-Guided Search Equations

Zhenxin Du, Dezhi Han, Guangzhong Liu, Kun Bi, Jianxin Jia

ABC elite, a novel artificial bee colony algorithm with elite-guided search equations, has been put forward recently, with relatively good performance compared with other variants of artificial bee colony (ABC) and some non-ABC methods. However, there still exist some drawbacks in ABC elite. Firstly, the elite solutions employ the same equation as ordinary solutions in the employed bee phase, which may easily result in low success rates for the elite solutions because of relatively large disturbance amplitudes. Secondly, the exploitation ability of ABC elite is still insufficient, especially in the latter half of the search process. To further improve the performance of ABC elite, two novel search equations have been proposed in this paper, the first of which is used in the employed bee phase for elite solutions to exploit valuable information of the current best solution, while the second is used in the onlooker bee phase to enhance the exploitation ability of ABC elite. In addition, in order to better balance exploitation and exploration, a parameter Po is introduced into the onlooker bee phase to decide which search equation is to be used, the existing search equation of ABC elite or a new search equation proposed in this paper. By combining the two novel search equations together with the new parameter Po, an improved ABC elite (IABC elite) algorithm is proposed. Based on experiments concerning 22 benchmark functions, IABC elite has been compared with some other state-of-the-art ABC variants, showing that IABC elite performs significantly better than ABC elite on solution quality, robustness, and convergence speed.