Dynamic Voronoi Treemaps: A Visualization Technique for Time-Varying Hierarchical Data


Treemaps are a widely used technique for the visualization of hierarchical data. In general, these techniques perform a space-filling recursive subdivision of a 2D space where the sizes of the created regions correspond to values of a particular data attribute. Several subdivision algorithms have been proposed to optimize specific criteria, such as region aspect ratio or stability. However these goals are often contradictory. For example, existing layout algorithms that optimize for aspect ratio–important for legibility–are typically not stable. For this reason, Treemaps are rarely used in animated displays of time-variant data. When they are applied to dynamic data sets, unstable Treemap layout algorithms produce poorly animated transitions that include discontinuous jumps in region position when values change. This paper introduces a new technique called Dynamic Voronoi Treemaps. Our layout algorithm is specifically designed to support smooth, real-time animation of time varying hierarchical data while maintaining desirable aspect ratios. We describe our novel approach and outline how it overcomes key limitations of prior Voroni-based Treemap work to enable the visualization of dynamic data. Results from an evaluation study of the technique are provided along with a brief use case highlighting a real-world application of Dynamic Voronoi Treemaps.

IBM Research Technical Report RC25132