Design of information visualization and case studies
Kang Zhang, Jie Hao · 2010
Information visualization (infovis) is the interdisciplinary study of visual representation of abstract information space. Infovis design refers to finding a visual metaphor which properly sheds light on the underlying data. The process of constructing an infovis technique is usually divided into the design stage, in which an infovis technique is designed, and the evaluation stage, in which the designed infovis technique has been implemented and then is enhanced according to the evaluators’ feedback. This dissertation provides a guideline called “5 Steps of Designing an Infovis Technique (5DITS)” for the design stage. In this dissertation, 5DITS is applied to design three infovis techniques. The first infovis technique, called Radial Edgeless Tree (RELT), is designed to visualize hierarchical information on devices with small screens. Few current hierarchy visualization techniques can achieve both desirable readability and efficient screen utilization. RELT solves this problem by recursively dividing a screen into non-overlapping polygons, each of which presents a node. Instead of edges between nodes, nodes’ relative positions are used to present connectivity. RELT has been successfully tested in stock market visualization and simulation of a cell phone browser. InfoShape, the second infovis technique, provides a global view for multi-dimensional information, which is organized in the table format. InfoShape consists of Record InfoShape (RInfoShape) and Dimension InfoShape (DInfoShape) that visualize multi-dimensional information as a 3D sphere whose appearance denotes how much the multi-dimensional information satisfies a pre-defined set of criteria. By comparing the shapes of different multiple information sets, global content similarities and differences can be quickly captured. RInfoShape and DInfoShape are evaluated on a Java program evaluation and US life table comparison, respectively. SciTrend, the last infovis technique, visualizes the popularity trend of several computer science research areas and relationships among them. The underlying data is a journal citation network which contains both hierarchical relation and additional non-hierarchical connectivity. The relationships between scientific area and research articles are hierarchical relationships. The citation relationships between research articles are non-hierarchical relationships. SciTrend provides static visual pattern and associated interactive functions, so that users can quickly capture the research areas’ popularity trends and relationships, and explore citation details among peer-reviewed journals.