Structure-based brushes: a mechanism for navigating hierarchicallyorganized data and information spaces

摘要

Interactive selection is a critical component in exploratorynvisualization, allowing users to isolate subsets of the displayedninformation for highlighting, deleting, analysis, or focusedninvestigation. Brushing, a popular method for implementing the selectionnprocess, has traditionally been performed in either screen space or datanspace. In this paper, we introduce an alternate, and potentiallynpowerful, mode of selection that we term structure-based brushing, fornselection in data sets with natural or imposed structure. Our initialnimplementation has focused on hierarchically structured data,nspecifically very large multivariate data sets structured vianhierarchical clustering and partitioning algorithms. The structure-basednbrush allows users to navigate hierarchies by specifying focal extentsnand level-of-detail on a visual representation of the structure.nProximity-based coloring, which maps similar colors to data that arenclosely related within the structure, helps convey both structuralnrelationships and anomalies. We describe the design and implementationnof our structure-based brushing tool. We also validate its usefulnessnusing two distinct hierarchical visualization techniques, namelynhierarchical parallel coordinates and tree-maps. Finally, we discussnrelationships between different classes of brushes and identify methodsnby which structure-based brushing could be extended to alternate datanstructures