|
As known, the artificial bee colony (ABC) algorithm is an optimization algorithm based on the intelligent foraging behavior of honey bee swarm that has been proven its efficacy and successfully applied to a large number of practical problems. Aiming at the trade-off between convergence speed and precocity of ABC algorithm with elite-guided search equations (ABC_elite), an enhanced version, namely EABC_elite, is proposed in this paper, and the improvements are twofold. As the global best (gbest) solution is introduced to the search equation and acceleration of the convergence in the bee phase of EABC_elite, the former in the ordinary solution is embodied to the search equation yet balance the gbest’s ability. The enhancement to the global search by making the information of gbest and ordinary solutions be adequately used while keeping the exploration–exploitation balance well maintained, the usual solution is introduced to the search equation to avoid the precocity problem in the onlooker bee phase of EABC_elite as the latter one. Experimental analysis and evaluations of EABC_elite against several state-of-the-art variants of the ABC algorithm demonstrate that the EABC_elite is significantly better than the compared algorithms in the feature selection problem. Also, the proposed EABC_elite algorithm is modified to combine the K -means initialization strategy and chaotic parameters strategy to further enhance the global search of EABC_elite for data clustering. Experimental results show that the derived EABC_elite clustering algorithm “Two-step EABC_elite,” TEABC_elite for short, delivered better and promising results than previous works for data clustering.
|
