A New Framework for Kmeans Algorithm by Combining the Dispersions of Clusters

Kmeans algorithm performs clustering by using a partitioning method which partition data into different clusters in such a way that similar object are present in one cluster that is within cluster compactness and dissimilar objects are present in different clusters that is between cluster separations. Many of the Kmeans type clustering algorithms considered only similarities among objects but do not consider dissimilarities. In existing system extended version of Kmeans algorithm is described. Both cluster compactness within cluster and cluster separations between clusters is considered in new clustering algorithm. Existing work initially developed a group of objective function for clustering and then rules for updating the algorithm are determined. The new algorithm with new objective function to solve the problem of cluster compactness within cluster and cluster separations between clusters has been proposed. Proposed FCS algorithm works simultaneously on both i.e. similarities among objects and dissimilarities among objects. It will give a better performance over existing kmeans.