Clustering based on density estimation using variable kernel and maximum entropy principle

Abstract

Clustering is one of the most important unsupervised classification strategies in data analysis. In this sense, a new clustering approach proposed a fast search algorithm of cluster centers based on their local densities has taken place. In the present paper, we suggest a new performed approach that combine the estimation of the local density and the use of the entropy. So the clustering algorithm is able to give automatically result without any iteration to optimize a cost function as the most popular clustering algorithm do, or either the user-interactive selection of the cluster centroids.