Information Technology Reference
In-Depth Information
C HAPTER T HIRTEEN
V ORONOI D IAGRAM B ASED D IMENSIONAL
A NCHOR A SSESSMENT FOR R ADIAL
V ISUALIZATIONS
A DAM R USSELL , K AREN D ANIELS
AND G EORGES G RINSTEIN
Abstract
This chapter explores the usefulness of Voronoi partitioning of a radial
visualization (RadViz) image space for selecting the most expressed
dimensions from high dimensional data sets. A metric is introduced for
assessing the effectiveness of this approach that links visualization quality
with selection and arrangement of dimensions.
I. Introduction
Researchers and data analysts working in various fields have a number of
tools and techniques at their disposal for visualizing and analysing their
data. RadViz and Parallel Coordinates are two well-known classes of data
visualizations that present high dimensional data sets in such a way that
the relationships between the data records and the dimensions are retained
and presented to the user. Daniels et al. [1] provide details of the invariant
properties of the data under the RadViz mapping. Many open problems
remain to be explored for this class of visualizations. This chapter focuses
on one such open problem: how do we quantify and compare the quality of
RadViz images?
RadViz is a 2D radial visualization that displays d dimensional data
within the unit circle. A distinctive feature of RadViz is the arranging of
labelled points which correspond to data dimensions placed on the
circumference of the unit circle. These labels are called “Dimensional
Search WWH ::




Custom Search