Multi-Level Area Balancing of Clustered Graphs

作     者：Wu, Hsiang-Yun Nollenburg, Martin Viola, Ivan 

作者机构：TU Wien A-1040 Vienna Austria King Abdullah Univ Sci & Technol KAUST Thuwal 23955 Saudi Arabia 

出 版 物：《IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS》 (IEEE Trans Visual Comput Graphics)

年 卷 期：2022年第28卷第7期

页      面：2682-2696页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0835[工学-软件工程] 

基　　金：European Union Vienna Science and Technology Fund (WWTF) [VRG11-010] Austrian Science Fund (FWF) [P31119] King Abdullah University of Science and Technology (KAUST) [BAS/1/1680-01-01] Austrian Science Fund (FWF) [P31119] Funding Source: Austrian Science Fund (FWF) Marie Curie Actions (MSCA) Funding Source: Marie Curie Actions (MSCA) 

主　　题：Layout Visualization Clustering algorithms Data visualization Shape Partitioning algorithms Chemical elements Graph drawing Voronoi tessellation multi-level spatially-efficient layout 

摘      要：We present a multi-level area balancing technique for laying out clustered graphs to facilitate a comprehensive understanding of the complex relationships that exist in various fields, such as life sciences and sociology. Clustered graphs are often used to model relationships that are accompanied by attribute-based grouping information. Such information is essential for robust data analysis, such as for the study of biological taxonomies or educational backgrounds. Hence, the ability to smartly arrange textual labels and packing graphs within a certain screen space is therefore desired to successfully convey the attribute data . Here we propose to hierarchically partition the input screen space using Voronoi tessellations in multiple levels of detail. In our method, the position of textual labels is guided by the blending of constrained forces and the forces derived from centroidal Voronoi cells. The proposed algorithm considers three main factors: (1) area balancing, (2) schematized space partitioning, and (3) hairball management. We primarily focus on area balancing, which aims to allocate a uniform area for each textual label in the diagram. We achieve this by first untangling a general graph to a clustered graph through textual label duplication, and then coupling with spanning-tree-like visual integration. We illustrate the feasibility of our approach with examples and then evaluate our method by comparing it with well-known conventional approaches and collecting feedback from domain experts.