Polygon-Based Spatial Clustering

Clustering geographic data using traditional methods often result in clusters that look dispersed over the geographic space and poorly reflect any underlying spatial structure. We propose a polygon-based spatial clustering approach, which models a spatial object as a polygon with three groups of attributes: general attributes, boundary attributes, and spatial events. We have developed a generalized distance function as a combination of distance functions defined on each group of attributes. The effectiveness of the approach is tested using a hydrological application. Experimental results show that this approach can organize the data into meaningful categories.