Mining Spatial Data via Clustering

Contributions from researchers in Knowledge Discovery are producing essential tools in order to better understand the typically large amounts of spatial data in Geographical Information Systems. Clustering techniques are proving to be valuable in providing exploratory analysis functionality while supporting approaches for automated pattern discovery in spatially referenced data and for the iden-tiication of important spatial relationships. However, there is little recognition of the broader context for which many clustering approaches are related. The concern for eecient ltering of outliers has divested attention from the actual problem being solved, which has resulted in a lack of recognition for the variety of approaches that could be modeled and in the re-discovery of solution approaches that are problematic or inferior. We present an overview of non-hierarchical clustering approaches for spatial data analysis. Our panoramic view to clustering large spatio-referenced datasets suggests that more eeective and eecient clustering methods are possible.