Our cluster-based hierarchical enhancements provide a multiresolutional view of the data and aid in revealing data trends at different degrees of summarizations. Our proximity-based coloring scheme assures that similar data clusters are shown in similar hue colors, while dissimilar ones in contrasting shades. The color scheme not only has a visual impact, it also aids in direct data selection by color. We augment our system with a set of navigation tools to support data localization and subspace drilling operations while maintaining context within the data space.
In summary, our system facilitates the interpretation of large, generic multivariate datasets through a systematic and interactive process involving aggregation, summarization, selection, and localization.
The ideas in this paper are implemented as OpenGL extensions to XmdvTool 3.1. As per tradition, the source code will be made available to the public domain (See http://www.cs.wpi.edu/ matt/research/XmdvTool).
This work is part of an ongoing research in WPI focusing on multivariate visualization of large datasets. Our future undertakings include extending the hierarchical methods to the other visualization modes in XmdvTool, including scatterplots, glyphs, and dimensional stacking. This will be done in a unified manner to maintain consistency across all modes. We are also investigating effective database management strategies within a large-scale multivariate visualization setting.
Acknowlegements
This work is sponsored by the NSF grant.