a wavelet-based clustering approach for spatial data(小波变换聚类).pdf

The VLDB Journal (2000) 8: 289–304 The VLDB Journal c Springer-Verlag 2000 WaveCluster: a wavelet-based clustering approach for spatial data in very large databases Gholamhosein Sheikholeslami, Surojit Chatterjee, Aidong Zhang Department of Computer Science and Engineering, State University of New York at Buffalo, Buffalo, NY 14260, USA Edited by S. Christodoulakis. Received June 9, 1998 / Accepted July 8, 1999 Abstract. Many applications require the management of automate the process of understanding spatial data by repre- spatial data in a multidimensional feature space. Cluster- senting the data in a concise manner and reorganizing spatial ing large spatial databases is an important problem, which databases to accommodate data semantics. It can be used in tries to find the densely populated regions in the feature many applications such as seismology (grouping earthquakes space to be used in data mining, knowledge discovery, or clustered along seismic faults), minefield detection (group- efficient information retrieval. A good clustering approach ing mines in a minefield), and astronomy (grouping stars in should be efficient and detect clusters of arbitrary shape. galaxies) [AF97, BR95]. It must be insensitive to the noise (outliers) and the order The aim of data-clustering methods is to group the ob- of input data. We propose WaveCluster, a novel clustering jects in spatial databases into meaningful subclasses. Due approach based on wa