A Modified Particle Swarm Optimization Using FCM for Moving Peaks Benchmark

Many optimization problems in real world are dynamic and they are changing over time. For resolving these problems, many different algorithms have been proposed. One of these, is PSO algorithm which has well supported its ability in resolving static problems. But this algorithm has some problems in dynamic environments. In this paper, an improved PSO algorithm with inertia parameter has been proposed for dynamic environments which increase the convergence speed of algorithm in getting close toward optimizations. In the proposed algorithm, in order to prevent excessive compression of groups at the end of each iteration, the distance between each group is measured and if this distance is lower than a threshold which is adjusted by a dynamic clustering, the worse group will be eliminated. When some changes is observed in the environment, first the particles’ memory is evaluated, then the particles are distributed inside a super globe with the best particle in the center to increase the group diversity. In order to optimize the results, a local search is applied around the best particle of the group. For evaluating the proposed algorithm, moving peaks benchmark was used. The findings showed that the proposed method operates better than other methods.