OST: a heuristic-based orthogonal partitioning algorithm for dynamic hierarchical data visualization

Abstract

Tools for intuitive visualization of dynamic datasets are highly demanded for capturing information and revealing potential patterns, especially in understanding the trend of data changes. We propose a novel resolution-independent heuristic algorithm, termed Orthogonal Stable Treemap (OST), to implicitly display dynamic hierarchical data value changes. OST adopts a site-based method as the Voronoi treemap (VT), to preserve the layout stability for diversified data values. Meanwhile, OST partitions the whole canvas with horizontal or vertical lines, instead of the lines with arbitrary orientations in VT. Technical innovations are made in three parts: Initialization of site state to speed up the algorithm and preserve the layout; efficient computation of orthogonal rectangular diagram to partition the empty canvas; self-adaption of site state to quickly reach an equilibrium. The performance of OST is quantitatively evaluated in terms of computation complexity, computation time, convergence rate, visibility, and stability. Moreover, qualitative evaluations (use case and user study) are demonstrated on the dynamic work-in-process dataset in the wafer fab. Evaluation results show that OST combines the advantages of layout stability and tidiness, contributing to easier and faster plot understanding.