Seyyed Masood Khademi - Department of Computer, Isfahan (Khorasgan) Branch, Islamic Azad University, Isfahan, Iran
Farsad Zamani Boroujeni - Department of Computer, Isfahan (Khorasgan) Branch, Islamic Azad University, Isfahan, Iran

Clustering process is the same data in the form of the cluster grouping. One of the techniques used in dataanalysis is cluster analysis, and one of the most popular clustering algorithms is k-means clustering. The initialcentroids generated randomly by the k-means algorithm that considers centroids great impact on the speed and accuracyof the final clusters. And also, the results of which are sensitive to the initial centroids. So the initial centroids for nonrandomshould be selected carefully. To resolve this, many researchers in this field determined to improve k-meansclustering. This paper introduces some methods of non-random selection of initial centroids, investigating theadvantages and disadvantages of them.

Clustering, Initial Centroids, k-means clustering, non-random centroids