NSS-AKmeans: An Agglomerative Fuzzy K-means clustering method with automatic selection of cluster number
Yanfeng Zhang, Xiaofei Xu, Yunming Ye · 2010
In this paper, we present a new Neighbor Sharing Selection based Agglomerative fuzzy K-means (NSS-AKmeans) algorithm for learning optimal number of clusters and generating better clustering results. The NSS-AKmeans can identify high density areas and determine initial cluster centers from these areas with a neighbor sharing selection method. To select initial cluster centers, we propose an agglomeration energy (AE) factor for representing global density relationship of objects, and a Neighbors Sharing Factor (NSF) for estimating local neighbor sharing relationship of objects. Then we use the Agglomerative Fuzzy k-means clustering algorithm to further merge these initial centers to obtain the preferred number of clusters and generate better clustering results. Experimental results on various data sets have shown that the NSS-AKmeans was very effective in automatically identifying the true cluster number as well as producing accurate clustering results.