Parallel Particle Swarm Optimization in Data Clustering
Particle Swarm Optimization (PSO) is a heuristic and population based optimization algorithm. PSO can be used
in clustering problems and dominates well known clustering algorithms such as K-Means and Fuzzy C-Means in the context of
not being stuck the local optima. In this study, a parallel PSO method was presented for clustering data.The parallel PSO was
tested with the iris and an artificial data set on a multiprocessor system. The results are compared to the sequential PSO in term
of fitness value, computation time and clustering success. The results show that parallel PSO outclasses sequential PSO in the
term of computation time in multiprocessor system. On the other hand, clustering success of parallel PSO is not less than
sequential PSO. Even its clustering success is %100 for the artificial dataset, whereas S-PSO’s clustering success is about %95.
Besides, parallel PSO converges to the optimum result for both datasets much earlier than S-PSO.
Index Terms — PSO, parallel computing, clustering, multiprocessor.