The synergistic combination of particle swarm optimization and fuzzy sets to design granular classifier

Abstract

Granulation extracts a bundle of similar patterns by decomposing universe. Hyperboxes are granular classifiers to confront the uncertainties in granular computing. This paper proposes a granular classifier to discover hyperboxes in three phases. The first phase of the proposed model uses the set calculus to build the hyperboxes; where, the means of the DBSCAN clustering algorithm constructs the structure. The second phase develops the geometry of hyperboxes to improve the classification rate. It uses the Particle Swarm Optimization (PSO) algorithm to optimize the seed-points and expand the hyperboxes. Finally, the third phase identifies the noise points; where, the patterns in the second phase did not belong to any hyperboxes. We have used the capability of membership function of a fuzzy set to improve the geometry of classifier. The performance of a proposed model is carried out in terms of coverage, misclassification error and accuracy. Experimental results reveal that the proposed model can adaptively choose an appropriate granularity.