A fuzzy particle swarm optimizer for unsupervised learning

We propose a hybrid procedure for the problem of unsupervised learning, based on a combination of fuzzy clustering and particle swarm optimization. Candidate solutions to this problem are considered by this procedure as particles that evolve in a swarm, where each particle is formed by a set of c prototypes representing the c clusters to be found in the learning database. The proposed method provides optimal solutions in terms of a fuzzy objective criterion. It is validated and compared to other methods using two benchmark datasets.