Particle swarm optimization with improved bio-inspired bees

Abstract

To solve difficult problems, bio-inspired techniques have been developed to find solutions. We will propose in this paper, a new hybrid algorithm called BEPSO (Bees Elitist Particle Swarm Optimization) which is an algorithm based on the PSO, and inspired by the behavior of bees when searching for food. BEPSO uses an iterative process where each iteration integrates a random search followed by an exploration phase and an intensification phase. Three parameters allow the user to control the duration of each phase which gives the algorithm the flexibility to adapt the different kind of problems. After each iteration, the best obtained result is stored, which gives BEPSO an elitist behavior. Besides, random fluctuation [1] avoids the premature convergence problem. Then we will present the tests on 25 well known Benchmark functions in the literature. We will compare BEPSO with six other variants of the PSO algorithm. The results have shown the superiority of our approach compared to the other approaches both in terms of quality of solutions and convergence speed.