Graphics Reference
In-Depth Information
AdatasetwithM items has M subsets, any one of which may be the one we re-
ally want. With a good data display, our own fantastic pattern-recognition abilities
can not only sort through this combinatorial explosion, but they can also extract in-
sightsfromthevisual patterns. hesearethecorereasons fordata visualization. With
parallel coordinates (abbrev.
-coords), the search for multivariate relations in high-
dimensional datasets is transformed into a -D pattern recognition problem. In this
chapter, the guidelines and strategy for knowledge discovery using parallel coordi-
nates are illustrated on various real datasets, one with variables froma manufac-
turing process. A geometric classification algorithm based on
-coords is presented
and applied to complex datasets. It has low computational complexity, providing the
classificationruleexplicitlyandvisually.heminimalsetofvariablesrequiredtostate
the rule are found and ordered by their predictive value. A visual economic model
of a real country is constructed and analyzed to illustrate how multivariate relations
can be modeled using hypersurfaces. he overview at the end provides a basic sum-
mary of
-coords and a prelude of what is on the way: the distillation of relational
information into patterns that eliminate need for polygonal lines altogether.
Introduction
14.1
Origins
14.1.1
Over half of our sensory neurons are devoted to vision, endowing us with tremen-
dous pattern recognition abilities. he goal of visualization is to couple this talent
to our problem-solving capabilities, in order to obtain insights from images. In or-
dertomakeiteasiertovisualize multivariate/multidimensional problems,numerous
mappings that visually encode multidimensional information in a two-dimensional
plane(seeFriendly, ;Tute, , , ),havebeeninvented.Suchmappings
augment our perception, which is limited by our habitation of a three-dimensional
world. Great successes in this field, like Minard's “Napoleon's March to Moscow,”
Snow's “dot map,” and others are, however, ad hoc (i.e., one-of-a-kind) and excep-
tional. Succinct multivariate relations are rarely made apparent by static displays. To
overcome this problem,wemust incorporatethe interactivity enabled bythetechno-
logical advances of the last half-century. In turn, this raises the issues of how to cre-
ate effective GUIs (graphical user interfaces), queries, exploration strategies, and of
course good information-preserving displays, but we are getting ahead of ourselves.
Legend has it that Archimedes, while constructing a proof, was so absorbed in
adiagramwhenhewaskilled byaRomansoldierthat hepleaded“Donotdisturbmy
circles” as he was being struck by a sword. Visualization first flourished in geometry
(which makes Archimedes' the first recorded death in defense of visualization). he
utilization of diagrams is strongly interwoven within the testing of conjectures and
the construction of proofs. his essence of visualization was eventually abstracted
Search WWH ::




Custom Search