An improved artificial bee colony algorithm for global numerical optimisation

Abstract

Artificial bee colony (ABC) is a simple and powerful optimisation algorithm, which simulates the random behaviour of honey bees. In this study, a modified version of ABC algorithm is proposed by considering: 1) initialising the population based on chaos theory; 2) utilising multiple searches, instead of single search, in employee and onlooker bee phases; 3) controlling the frequency of perturbation by a modification rate. The proposed algorithm implemented by C# programming language and executed on benchmark functions of Sphere, Rosenbrock, Rastrigin, non-continuous Rastrigin, Griewank, Schwefel, Schwefel 1.2, Schwefel 2.2, Step and Ackley. The performance of proposed ABC algorithm is compared with ABC and modified ABC algorithms, on different tests, according to measures of mean and standard deviation. Findings showed the suggested algorithm outperforms ABC and modified ABC algorithms in 65% of test cases in getting best mean and in 55% of cases for standard deviation.