ClusterSets: Optimizing Planar Clusters in Categorical Point Data

In geographic data analysis, one is often given point of different categories (such as facilities a university categorized by department). Drawing upon recent research on set visualization, we want to visualize category membership connecting points the same with visual links. Existing approaches that follow this path usually insist all members category, which may lead many crossings and clutter. We propose an approach avoids between connections completely. Instead subdivide into smaller, local clusters where needed. do case study comparing legibility drawings produced our those existing approaches. problem formulation, are additionally graph G whose edges express some sort proximity. Our aim find subgraph G′ following properties: (i) connect only (ii) no two cross, (iii) number connected components (clusters) minimized. then in G′. For arbitrary graphs, resulting optimization problem, Cluster Minimization, NP-hard (even approximate). Therefore, introduce heuristics. extensive benchmark test real-world data. Comparisons exact solutions indicate heuristics astonishing well for certain relative-neighborhood graphs.