K-Means Clustering - Review of Various Methods for Initial Selection of Centroids
Sunil Kumar · 2013
Organising data into sensible groups is the most fundamental way of understanding and learning. Clustering helps to organise data based on natural grouping without any category labels to identify the clusters.One of the most popular and simplest partitional clustering algorithms is the K -Means published in 1955. K -Means algorithm is computationally expensive and insists the selection of number of clusters initially. The final clusters depend entirely on the initial selection of centroids. Several modifications have been proposed for the K -Means clustering method. Some such proposals are summarised and reviewed with experimental results.