A new algorithm to get the initial centroids

We investigate in This work the standard k-means clustering algorithm and give our improved version by selecting better initial centroids that the algorithm begins with. First we evaluate the distances between every pair of data-points; then try to find out those data-points which are similar; and finally construct initial centroids according to these found data-points. Different initial centroids lead to different results. If we can find initial centroids which are consistent with the distribution of data, the better clustering can be obtained. According to our experimental results, the improved k-means clustering algorithm has the accuracy higher than the original one.