An ILP Approach to Spatial Clustering

Spatial clustering is a fundamental task in Spatial Data Mining where the goal is to group nearby sites and form clusters of homogeneous regions. Spatial clustering must be driven by the discrete spatial structure of data that expresses the (spatial) relational constraints between separate sites. Only similar sites (transitively) connected in the discrete spatial structure may be clustered together. CORSO is a novel clustering method that resorts to an ILP approach to mine spatial data and exploits the concept of neighborhood to capture relational constraints. Spatial data are expressed in a first-order formalism and similarity among structured objects is computed as degree of matching with respect to a common generalization. Two real-world applications are described.