Improving K-Mean Method by Finding Initial Centroid Points

The paper is concerned with Improving k-Mean Algorithm in terms of accuracy by selecting the best initial seed points based on the provided k value. This paper presents two modified k-mean method for the selection of initial centroid points. In the first method based on the calculated k value with the help of elbow method, the original sorted data based on distances calculated using Euclidean distance method is divided into k equal partitions. And the mean of each partition is considered as initial centroid points. And in the second method the number of k is chosen randomly and the mean of each partition is considered as initial centroid points. We compared within cluster distance and number of iterations. Modified k-mean methods are better than original k-mean method as the distance within the clusters are less in modified k-mean than the original k-mean and the accuracy is also better.