Enhancing the particle swarm optimizer via proper parameters selection

Unlike many other computational intelligence techniques, the particle swarm optimizer (PSO) has few parameters to tune. However, properly chosen values for these parameters can positively affect the accuracy of the obtained results as well as the time consumed during the search process. Many parameters have been added to the originally developed PSO to modify or to improve the performance of the technique but yet, the swarm size, number of iterations and individuals flying velocities are still the most dominant parameters. The paper examines the PSO´s parameters, describes their characteristics and provides guidelines for determining values for these parameters. A quick statistical experiment is used to fine-tune these parameters for the class of constrained optimization problem considered. The results show that the particle swarm optimizer is quite robust and provides good solution for reasonable choice of the values of the parameters within fairly wide range.