A Qualified Search Strategy with Artificial Bee Colony Algorithm for Continuous Optimization

Abstract

One of the most popular population-based and swarm intelligence algorithms is the artificial bee colony. Although the ABC method is known for its efficiency in exploration, it has a poor performance in exploitation ability. It uses a single solution search equation that does not provide a balance between exploration and intensification adequately, and this is one of the most common problems in optimization techniques. This study proposes an artificial bee colony algorithm with a qualified search strategy (QSSABC) that uses four different solution search equations to deal with these problems. In order to increase the ability of exploitation, the QSSABC uses the global best solution of population in both of these equations. Equations in the QSSABC method are selected by roulette-wheel method based on their success rates, and equation with the lowest success rate within determined periods is eliminated. The equations’ success rates are reset at the end of each period, and it is expected that equations will prove their success again in every period. This qualified search strategy ensures an efficient use of number of function evaluations, and also it achieves balance between global and local search. To evaluate accuracy and performance of the QSSABC, twenty-eight classical functions, twenty-four CEC05 functions and thirty CEC14 functions were used. Simulation results showed that the QSSABC surpasses other methods such as distABC, MABC, ABCVSS in classical functions, and that it is a successful tool for problems with different characteristics by showing better performance over other methods for CEC05 and CEC14 test functions.