An individual dependent multi-colony artificial bee colony algorithm

Artificial bee colony (ABC) is a well-known swarm intelligence based algorithm that simulates the foraging behavior of honey bees for food sources. However, the basic ABC only evolves one colony and both the employed bee phase and onlooker phase utilize the same solution search equation, which performs well in exploration but poorly in exploitation. Inspired by the good working efficiency of labor division and the coordination of complementary elements, we propose a novel individual dependent multi-colony ABC algorithm, abbreviated as IDABC, in which the whole colony is divided into three sub-colonies, i.e., inferior sub-colony, mid sub-colony and superior sub-colony, based on the fitness function values of the individuals involved. Three evolution operators with different searching biases are introduced into the corresponding sub-colonies in order to play different roles. Furthermore, we improve the related evolution operators by incorporating the fitness and distance information of individuals and the control parameters involved in the food source perturbation that are dynamically adjusted according to the search experience. And an orthogonal learning mechanism is suggested for scout bee searching so as to generate potential solutions. The proposed IDABC is examined in a set of benchmark instances taken from the CEC2013 and CEC2014 competition, and comparative results demonstrate the competitive performance of IDABC.