Comparative analysis of selection schemes used in artificial bee colony algorithm

Abstract

Artificial bee colony (ABC) algorithm, which has explicit strategies to balance intensification and diversification, is a simple and effective swarm intelligence algorithm for both continuous and combinatorial optimisation problems. Selection scheme, which is used by onlooker bees to select employed bees to follow, is an important factor for ABC algorithm to keep good balance between intensification and diversification. In this paper, a normalised fitness function for minimisation problem is first proposed for roulette wheel selection, so ABC can still have enough selection pressure in late stage. And then, aim to analyse the effect of different selection schemes, systematic experiments were carried for roulette wheel selection, rank selection, tournament selection, and disruptive selection on a set of test functions with different dimensions and different computation resources. Simulation results show that the suitability of a selection scheme depends not only on the features of test functions and but also on the computation resources used. No selection scheme can always outperform other selection schemes on all test functions.