Fuzzy cluster analysis of spatio-temporal data

Abstract

The quantities of earth science data collected have necessitated the development of new data mining tools and techniques. Mining this data can produce new insights into weather, climatological and environmental trends that have significance both scientifically and practically. This paper discusses the challenges posed by earth science databases and examines the use of fuzzy K-Means clustering for analyzing such data. It proposes the extension of the fuzzy K-Means clustering algorithm to account for the spatio-temporal nature of such data. The paper introduces an unsupervised fuzzy clustering algorithm, based on the fuzzy K-Means and defines a cluster validity index which is used to determine an optimal number of clusters. It is shown experimentally that the algorithms are able to identify and preserve regions of meteorological interest. © Springer-Verlag Berlin Heidelberg 2003.