The Dissimilated Krill Herd Algorithm with Aggregation Degree

In view of shortcomings of the basic krill herd algorithm (KH)——easily trapping in local optimization and low search accuracy, a dissimilated krill herd algorithm with aggregation degree is proposed in this paper. According to the variation of concentration (which plays as the diversity criterion of population), the algorithm increases the krill's evolving diversity by a selection strategy on two opposite directions, moreover, a random number strategy is introduced to simulate the external disturbance of krill, which actually takes the place of the influence of stochastic diffusion movement of original krill herd algorithm, and then the average distance index is used to increase the dissimilation probability of local search. Also the proposed algorithm takes the reverse of optimal position as the dissimilation search direction to expand the search space of the group, so that it can ensure a satisfied the global search ability. The experimental results show that the dissimilation strategy can gain the promising availabilities of the optimization accuracy and the global search ability.