An Improved Artificial Bee Colony Algorithm for Balancing Local and Global Search Behaviors in Continuous Optimization

Abstract

The artificial bee colony, ABC for short, algorithm is population-based iterative optimization algorithm proposed for solving the optimization problems with continuously-structured solution space. Although ABC has been equipped with powerful global search capability, this capability can cause poor intensification on found solutions and slow convergence problem. The occurrence of these issues is originated from the search equations proposed for employed and onlooker bees, which only updates one decision variable at each trial. In order to address these drawbacks of the basic ABC algorithm, we introduce six search equations for the algorithm and three of them are used by employed bees and the rest of equations are used by onlooker bees. Moreover, each onlooker agent can modify three dimensions or decision variables of a food source at each attempt, which represents a possible solution for the optimization problems. The proposed variant of ABC algorithm is applied to solve basic, CEC2005, CEC2014 and CEC2015 benchmark functions. The obtained results are compared with results of the state-of-art variants of the basic ABC algorithm, artificial algae algorithm, particle swarm optimization algorithm and its variants, gravitation search algorithm and its variants and etc. Comparisons are conducted for measurement of the solution quality, robustness and convergence characteristics of the algorithms. The obtained results and comparisons show the experimentally validation of the proposed ABC variant and success in solving the continuous optimization problems dealt with the study.