QuestVis and MDSteer : the visualization of high-dimensional environmental sustainability data

• Main Author: Williams, Matt
• Format: Others
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/15921

The visualization of large high-dimensional datasets is an active topic within the research area of information visualization (infovis), a research area that studies the visual representations of complex abstract datasets. My thesis presents two infovis systems that were motivated by the desire to explore a 294-dimensional environmental sustainability dataset. Our collaborators developed the environmental dataset from expert knowledge on ecological, economical, and social systems which were used to model future scenarios consisting of 294 measures of environmental sustainability such as urban population, water supply levels, or tonnes of waste. Since these complex systems and large datasets are difficult for a non-expert user to comprehend, we developed QuestVis, a tool that applies infovis theories and techniques to improve the comprehensibility during exploration of the environmental dataset. The tool consists of three components: the input panel, the Multiscale Dimension Visualizer (MDV), and the Scenario Space Explorer (SSE). The MDV presents up to ten 294-dimensional future scenarios simultaneously on the screen to enable users to get a quick overview of the data. The simultaneous presentation also enables users to compare multiple future scenarios side-by-side. The SSE presents the space of all 120 000 future scenarios in an interactive two-dimensional layout which provides the user an overview of the possibilities. The SSE is tightly coupled with the MDV to provide context to the specific future scenarios that are presented in the MDV. These tightly linked components together provide an overview-hdetails framework within which users can effectively explore the dataset and immediately see the consequences of their choices. The creation of the dimensionality reduced overview in QuestVis led to a second research direction. We realized that current implementations of Multidimensional Scaling (MDS), a technique that attempts to best represent data point similarity in a low-dimensional embedding, are not suited for many of today's largescale datasets. This realization motivated us to develop MDSteer, a steerable MDS computation engine and visualization tool that progressively computes an MDS layout and handles datasets of over one million points. Our technique employs hierarchical data structures and progressive layouts that allow the user to steer the computation of the algorithm to the interesting areas of the dataset. The algorithm iteratively alternates between a layout stage in which a sub-selection of points are added to the set of active points affected by the MDS iteration, and a binning stage which increases the depth of the bin hierarchy and organizes the currently unplaced points into separate spatial regions. This binning strategy allows the user to select onscreen regions of the layout to focus the MDS computation into the areas of the dataset that are assigned to the selected bins. We show both real and common synthetic benchmark datasets with dimensionalities ranging from 3 to 300 and cardinalities of over one million points. === Science, Faculty of === Computer Science, Department of === Graduate