Abstract: To deal with the problem that classical k-means algorithm inefficiently adapt to clustering for all kinds of clusters,in this paper an algorithm which is improved on k-means algorithm using optimization cluster-center was proposed.It divides large or extended-shaped cluster into a number of globular clusters,and then merges these small clusters.Firstly,a group of cluster centers located in the high-density region are selected,and the object around the cluster center is divided to its nearest cluster center forming the sub-cluster.Then the merger is completed in accordance with sub-cluster connectivity between sub-clusters.Experimental results show that the algorithm can adapt to irregular shape cluster and is simple.

Key words: Clustering algorithm,Cluster-center,Cluster consolidation,Fast