Transcript An Extension of Table Lens

An Extension of Table Lens
CPSC 533 Information Visualization
Course Project, Term 2, 2003
Fengdong Du
Table Lens Technique

Show a large amount of information in a
relatively small table.
 Preserve Global context
 Detail-on-demand presentation
 Support simple pattern discovery
Extension of Table Lens

Display not only large amount of data but
also relatively large dimensionality.
 Combine data mining techniques to
facilitate discovering more complicated
pattern.
Project Proposal: Combine Table Len
with Classification Rule Mining
Classification Mining

Generate classification rules given a set of
training data.
 Class label is treated as a function of a set
of non-class attribute.
 Find the minimum set of attributes that
predict the class attribute with high
accuracy.
Example
PlayTennis Outlook
Windy Temp.
Yes
No
Overcast
Raining
False
True
Normal
Normal
…
…
…
…
Rule: Outlook=overcast
PlayTennis=Yes
Combining Table Len with
Classification Rule Mining

Put class label attribute and class predict as
the most interested attribute.
 Class attribute and class predict are never
demagnified.
(Continued)

All the remaining attributes are demagnified
by their “importance” of classing the data.
 Possibly show a relatively large
dimensionality, e.g. less than 50 attributes.
(Continued)

Attribute details are showed when users
move focus to head of that column.
 Data record details are showed when users
move focus to that row.
(subject to a lot of change and improvement)