A Botanically Inspired High-Dimensional Visualization with Multivariate Glyphs

Abstract

It is difficult for the average viewer to assimilate and comprehend huge amounts of high-dimensional data. It is important to present data in a way that allows the user a high level understanding of the overall organization and structure without losing the ability to study low level detail as needed. Although hierarchically clustered data is already organized, many current means of presenting such data give the user little more than an overview of the organization. It would be useful to see more information about the data even at a high level and to examine specific clusters as needed. We want to understand the relationships of the clusters in terms of the underlying data, and to understand the extent and variability of the data without requiring examination of each data item. To meet these goals, we present an aesthetically appealing visualization based on botanical trees which preserves the natural order of hierarchically organized data. Hierarchical data is rendered as a simple branched tree. The tree gives an overview of the relationships among various clusters and is supplemented with two glyphs which allow the user to focus in on specific clusters of the data at different levels of detail. At a medium level of focus, a cluster glyph based on a radial, space filling approach shows the subtree rooted at a specified cluster. At a low level of detail, the branch glyph allows the viewer to see not only aggregate information about the cluster but the extent and variability of the component clusters.