A Stage by Stage Pruning Algorithm for Detecting the Number of Clusters in a Dataset

Determining the number of clusters in a dataset has been one of the most challenging problems in clustering analysis In this paper we propose a stage by stage pruning algorithm to detect the cluster number for a dataset The main idea is that from the dataset we can search for the representative points of clusters with the highest accumulation density and delete the other points from their neighborhoods stage by stage As the radius of the neighborhood increases the number of searched representative points decreases However, the structure of actual clusters of the dataset makes the representative point number be stable at the true cluster number in a relative large interval of the radius, which helps us to detect the cluster number It is demonstrated by the simulation and practical experiments that the proposed algorithm can lead to an effective estimate of the cluster number for a general datasethttp://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000284034500028&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=8e1609b174ce4e31116a60747a720701Computer Science, Artificial IntelligenceComputer Science, Information SystemsComputer Science, Theory & MethodsEICPCI-S(ISTP)