New Genetic Gravitational Search approach for Data Clustering using K-Harmonic Means

Abstract

this article the new hybrid data clustering approach, Gravitational Genetic KHM, based on Genetic algorithm (GA), Gravitational Search Algorithm (GSA) and K-harmonic Means (KHM) is proposed. Data Clustering is used to group similar set of objects into set of disjoint classes, object in class are highly similar than the objects in other classes. Among various clustering methods, KHM is one of the most popular clustering techniques. KHM is applied widely and works well in many fields, but this method runs in local optima. In the proposed approach the merits of Genetic Algorithm are used to escape the KHM clustering from local optima and to overcome the slow convergence speed of GSA. This paper is presented as work-in-progress in which the work model is proposed and some intermediate results are discussed which in turn will be compared with existing hybrid algorithms. The results are tested on several datasets. Keywords-Harmonic Means (KHM), Clustering, Gravitational Search Algorithm (GSA), Genetic Algorithm (GA).