A novel hybrid binary PSO algorithm

Abstract

The continuous PSO algorithm has been widely researched and also applied as an intelligent computational technique to solve problems requiring iterative solutions based on some predefined objective function. However, the research on binary version of PSO (DBPSO) is still underway. The major research concerns are to accelerate the convergence speed retaining the search ability and reliability of the algorithm. To achieve this, a novel hybrid binary particle swarm optimization (HBPSO) algorithm is proposed in this paper. It combines the PSO's concept and GA. In the existing standard binary PSO (DBPSO) two new operators such as crossover and mutation are incorporated to accelerate the convergence speed and to avoid possible stuck in local optimum thereby maintaining population diversity. The proposed HBPSO algorithm has been studied on 6 bench mark optimization problems. The experimental results such as minimum fitness, mean fitness, and variance of fitness over 50 consecutive trials on each objective function indicate that the HBPSO algorithm consistently outperforms the DBPSO and its variants in terms of convergence speed and search accuracy on a bulk of bench mark problems. © 2011 Springer-Verlag.