International Science and Technology Journal

Clustering is considered as a significant technique of classifying various groups of datasets. In this technique, the similar data objects are grouped together relocated in a new cluster .For this purpose, The K-means algorithm Cluster analysis is used .Yet, it still has several disadvantages: First, attaining the appropriate solutions. Second, results significantly affected by the selection of initial centroid points .Third, number of clusters has to be identified in advance. So the initial centroid for clustering is a key factor in k-Means. In this paper, a selection of points as initial centroids based on the distance is used instead of the random selection procedure. Precisely, the furthest apart points were selected for such a purpose .This proposed method has achieved better clustering process .Consequently, the number of repetitions is decreased, and the optimal solution is achieved