GRASP IN THE RESOLUTION OF THE CLUSTERING PROBLEM

Abstract

The clustering could be approached as a combinatorial optimization problem when the clusters are a partition of an objects set. The Grasp meta-heuristic is a relatively recent technic that had been used to solve of an e_cient manner several combinatorial optimization problems. In this work, we adapted the Grasp metaheuristic to solve the clustering problem based on the basis of K-Means algorithm. The proposed algorithm, named GraspKM, takes advantages of fast convergence of K-Means algorithm avoiding the inconvenience of obtaining a local optimal. The algorithm shows to be better than KMeans algorithm and it is comparable with another meta-heuristic method reviewed with respect to e_ciency. The computational experiments had been realized with a data collection extensively used on clustering literature.