Dingding Wang, Florida International University, USA
Chris Ding, University of Texas at Arlington, USA
Tao Li, Florida International University, USA

Links

Session:
Springer Link:

Abstract

The widely used K-means clustering deals with ball-shaped (spherical Gaussian) clusters. In this paper, we extend the K-means clustering to accommodate extended clusters in subspaces, such as line-shaped clusters, plane-shaped clusters, and ball-shaped clusters. The algorithm retains much of the K-means clustering flavors: easy to implement and fast to converge. A model selection procedure is incorporated to determine the cluster shape. As a result, our algorithm can recognize a wide range of subspace clusters studied in various literatures, and also the global ball-shaped clusters (living in all dimensions). We carry extensive experiments on both synthetic and real-world datasets, and the results demonstrate the effectiveness of our algorithm.