Authors:
René Rosenbaum
1
;
Daniel Engel
2
;
James Mouradian
1
;
Hans Hagen
2
and
Bernd Hamann
3
Affiliations:
1
University of California, United States
;
2
University of Kaiserslautern, Germany
;
3
University of California and University of Kaiserslautern, United States
Keyword(s):
High-dimensional Data, Projections, Value Visualization, Relation Visualization, Interaction, Scalability.
Related
Ontology
Subjects/Areas/Topics:
Abstract Data Visualization
;
Computer Vision, Visualization and Computer Graphics
;
General Data Visualization
;
Information and Scientific Visualization
;
Visual Representation and Interaction
;
Visualization Algorithms and Technologies
Abstract:
Structural Decomposition Trees (SDTs) have been proposed as a completely novel display approach to tackling the research problem of visualizing high-dimensional data. SDTs merge the two distinct classes of relation and value visualizations into a single integrated strategy. The method is promising; however, statements regarding its meaningful application are still missing, constraining its broad adoption. This paper introduces solutions for still-existing issues in the application of SDTs with regard to interpretation, interaction, and scalability. SDTs provide a well-designed initial projection of the data to meaningfully represent its properties, but not much is known about how to interpret this projection. We are able to derive the data’s properties from their initial representation. The provided methods are valid not only for SDTs, but also for projections based on principal components analysis, addressing a frequent problem when applying this technology. We further show how inte
ractive exploration based on SDTs can be applied to visual cluster analysis as one of its application domains. To address the urgent need to analyze vast and complex amounts of data, we also introduce means for scalable processing and representation. Given the importance and broader relevance of the discussed problem domains, this paper justifies and further motivates the usefulness and wide applicability of SDTs as a novel visualization approach for high-dimensional data.
(More)