Geospatial Cluster Tessellation Through the Complete Order- k Voronoi Diagrams

In this paper, we propose a postclustering process that robustly computes cluster regions at different levels of granularity through the complete Order-k Voronoi diagrams. The robustness and flexibility of the proposed method overcome the application-dependency and rigidity of traditional approaches. The proposed cluster tessellation method robustly models monotonic and nonmonotonic cluster growth, and provides fuzzy membership in order to represent indeterminacy of cluster regions. It enables the user to explore cluster structures hidden in a dataset in various scenarios and supports “what-if” and “what-happen” analysis. Tessellated clusters can be effectively used for cluster reasoning and concept learning.