Artificial bee colony algorithm with distribution-based update rule

Abstract

In last decades, lots of nature-inspired optimization algorithms are developed and presented to the literature for solving optimization problems. Generally, these optimization algorithms can be grouped into two categories: evolutionary algorithms and swarm intelligence methods. Evolutionary methods try to improve the candidate solutions (chromosomes) using evolutionary operators such as crossover, mutation. The methods in swarm intelligence category use differential position update rules for obtaining new candidate solutions. The popularity of the swarm intelligence methods has grown since 1990s due to their simplicity, easy adaptation to the problem and effectiveness in solving the nonlinear optimization problems. One of the popular members of swarm intelligence algorithms is artificial bee colony (ABC) algorithm which simulates the intelligent behaviors of real honey bees and uses differential position update rule. When food sources which present possible solutions for the optimization problems gather on the similar points within the search space, differential position update rule can cause a stagnation behavior in the algorithm during the search process. In this paper, a distribution-based solution update rule is proposed for the basic ABC algorithm instead of differential update rule to overcome stagnation behavior of the algorithm. Distribution-based update rule uses the mean and standard deviation of the selected two food sources to obtain a new candidate solution without using any differential-based processes. This approach is therefore prevents the stagnation in the population. The proposed approach is tested on 18 benchmark functions with different characteristics and compared with the basic variants of ABC algorithm and some nature-inspired methods. The experimental results show that the proposed approach produces acceptable and comparable solutions for the numeric problems.