# Data Visualization in Data Mining by qfK0DH

VIEWS: 9 PAGES: 24

• pg 1
```									Information Visualization
in Data Mining
S.T. Balke
Department of Chemical Engineering
and Applied Chemistry
University of Toronto
Motivation

   Data visualization
– relies primarily on human cognition for value
discovery;
– permits direct incorporation of human ingenuity
and analytic capabilities into data mining;
– can very effectively deal with very large
quantities of data;
– powerfully combines with machine-based
discovery techniques.
Uses

   Explorative Analysis
– Data cleaning
– Provide hypotheses
   Confirmative Analysis
– Confirm or reject hypotheses
   Presentation
http://www.alz.washington.edu/DATA2001/GERALD1/sld011.htm
Calculated Properties of
the Anscombe Data Sets

mean of the x values = 9.0

mean of the y values = 7.5

equation of the least-
squared regression line is: y
= 3 + 0.5x

sums of squared errors
Calculated Properties of
the Anscombe Data Sets

regression sums of squared errors
(variance accounted for by x) = 27.5

residual sums of squared errors (about
the regression line) = 13.75

correlation coefficient = 0.82

coefficient of determination = 0.67
The Anscombe Data
Marley, 1885
Snow’s Cholera
Map, 1855
http://pupgg.princeton.edu/disk20/anonymous/groth/lick/licknorth.gif
Graphical Excellence
(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)

Graphical displays should:
 show the data
 induce the viewer to think about the substance, not the
methodology
 avoid distorting what the data says
 present many numbers in a small space
 make large data sets coherent
 encourage the eye to compare different pieces of data
 reveal the data at several levels of detail (broad overview to
fine structure)
 serve a reasonably clear purpose: description, exploration,
tabulation, or decoration
 be closely integrated with the statistical and verbal
descriptions of the data set.
Graphical Excellence

   Gives the viewer the greatest number
of ideas in the shortest time with the
least ink in the smallest space.
   Nearly always multivariate.
   Requires telling the truth about the
data.

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Lie Factor=14.8

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Lie Factor

size of effect shown in graphic
Lie Factor 
size of effect in data

(27.5  18.0)100
Lie Factor         18         14.8
(5.3  0.6)100
0.6

Require: 0.95<Lie Factor<1.05

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Using Area for One
Dimensional Data

Lie Factor=2.8

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
More guidelines:

   The number of information-carrying
(variable) dimensions depicted should
not exceed the number of dimensions
in the data.
   No legends: use labels on graph
   Graphics must not quote data out of
context.
(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Data Ink Ratio

data ink
Data ink Ratio 
total ink used to pr int the graphic

Data ink Ratio = proportion of a graphic’s ink devoted to the

non-redundant display of data-information.

Data ink Ratio=1.0-(proportion of a graphic that can be erased
without loss of data-information)

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Maximize Data Density

number of entries in the data matrix
data density of a graphic 
area of data graphic

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
Beware Chartjunk

NO

“Isn’t it remarkable that the computer can be programmed
to draw like that.”

YES:

“My, what interesting data!”

(E.R. Tufte, “The Visual Display of Quantitative Information”, 2nd edition)
How to Say Nothing with
Information Visualization
http://www.crs4.it/~zip/13ways.html

   Never include a color legend.
   Avoid annotation.
   Never mention error characteristics of the
visualization method.
   When in doubt, smooth.
   Don’t say how long it required to plot.
   Never compare your results with other data
visualization techniques.
   Never cite references for the data.
   Claim generality but show results from a single data
set.
   Use viewing angle to hide blemishes in 3D objects.
An Overview of
Information Visualization
Methods
http://www.informatik.uni-
halle.de/~keim/tutorials.html
Methods of Interest

   Scatterplot Matrices
   Parallel Coordinates
   Pixel Oriented Methods
   Icon based Methods
   Dimensional Stacking
   Treemap
Assignment 1: see
handout
Some websites of
interest:
   http://dmoz.org/Computers/Software/Databases/Data_Mining/
Public_Domain_Software/
   http://www.cs.man.ac.uk/~ngg/InfoViz/Projects_and_Products/
Visualization/

Try a search at google.com using the
followng key words together: