visualization tree multiple linked analytical decisions
DESCRIPTION
http://www.icmc.usp.br/~junio/PublishedPapers/RodriguesJr_et_at%20-SmartGraphics-VisualizationTree.pdfTRANSCRIPT
The 5th International Symposium on Smart Graphics(SG-2005) - Frauenwoerth Cloister, Germany, August 22-24, 2005
Visualization Tree, multiple Visualization Tree, multiple linked analytical decisionslinked analytical decisions
Rodrigues Jr., José Fernando
Traina, Agma J. M.
Traina Jr., Caetano
University of São PauloComputer Science Department
ICMC-USPBrazil
(SG-05)
2/30
• Introduction
• Developed System
• Interaction Systematization
• Summarization Features
• Conclusions and Ending
Out
line
Out
line
• Introduction
(SG-05)
3/30
Techniques used to generate scenes whose graphical attributes rely on the data values
Info
vis
Info
vis Information Visualization (Infovis), manages
to develop techniques for the analysis of data sets that do not have an intrinsic graphical nature
Visualization
Information Visualization(InfoVis)
Scientific Visualization
(SG-05)
4/30
Increasing volume of data that cannot be well utilized to produce useful knowledge
Raw visualization techniques are limited in the task of data analysis, while datasets are unlimited both in size and complexity
The efficient analysis of multivariate data can provide assistance in decision making
There is a need for visualization mechanisms that reduce the
drawback of massive datasets.
Mot
ivat
ion
Mot
ivat
ion
(SG-05)
5/30
The
pro
blem
The
pro
blem
Due to overlap of graphical items, some regions of the visualization seam like blots in the display
Massively populated datasets tend to result in a visualization scene with an unacceptable level of clutter
Overlap of graphical items
Visual clutter
(SG-05)
6/30
• Introduction
• Developed System
• Interaction Systematization
• Summarization Features
• Conclusions and Ending
Out
line
Out
line
• Introduction
• Developed System
(SG-05)
7/30
Dev
elop
ed s
yste
mD
evel
oped
sys
tem
Visualization Tree– Data analysis multiple visualization techniques
– Graphical overlap Visual pipeline
– Cognitive flow Workspaces refinement and composition
– Visual clutter Tree scheme
– Enhance exploration New interaction systematization based on the tree metaphor
– Overpopulated data sets Frequency plot
– Data summarization Statistics presentation
– Hypothesis formulation Relevance plot
The Visualization Tree system is a systematic effort to enhance the
InfoVis practice by utilizing integrated presentation, interaction
and summarization mechanisms.
(SG-05)
8/30
• Introduction
•Developed System
• Interaction Systematization
• Summarization Features
• Conclusions and Ending
Out
line
Out
line
• Introduction
• Developed System
• Interaction Systematization
(SG-05)
9/30
Mul
tipl
e te
chni
ques
Mul
tipl
e te
chni
ques
Multiple visualization techniques at each workspace permits to explore each technique’s advantages in order to aid the analysis process
Scatter Plots
Parallel Coordinates
Star Coordinates
Table Lens
(SG-05)
10/30
Vis
ual p
ipel
ine
Vis
ual p
ipel
ine
The visual pipeline allows to extend one workspace’s visualization to multiple workspaces
It naturally diminishes graphical items overlap by extending the boundaries in derived workspaces
Via successive pipelines, it is possible to grasp details until only one item
populates its own workspace.
(SG-05)
11/30
Tre
e sc
hem
eT
ree
sche
me
The tree scheme allows to build multiple views in a decision-tree style
Cars
European
Japanese
American
4 cylinders
3 cyliners
1976 - 1982
1970 - 1976
(SG-05)
12/30
Tre
e sc
hem
eT
ree
sche
me
The use of multiple views is a known strategy that can help to diminish user cognitive overhead:– single views create cognitive overhead by
requiring simultaneously comprehension of diverse data
– easier to accomplish than single view memory-based comparison
– “divide and conquer,” aiding memory by reducing the amount of data they need to consider at the same time
In other words, the tree scheme can help to bypass the drawbacks of
visual clutter presentation.
(SG-05)
13/30Cco
mpo
siti
onC
com
posi
tion
Besides refining the views, it may be interesting to merge views for extra analytical possibilities:
– when two or more views have similar, correlated or worthy-comparing semantics
– for easy comparison, it may be worthy to put together branches of the tree in side-by-side, rather than node-like, positioning
Cars
European
Japanese
American
4 cylinders
3 cyliners
1976 - 1982
1970 - 1975
(European 4 cylinder moels) OR
(Older American models)
The composition of workspaces addresses these issues in an easy-to-
use interaction.
(SG-05)
14/30Inte
ract
ion
Inte
ract
ion
syst
emat
izat
ion
syst
emat
izat
ion
The developed system proposes a new interaction systematization to explore multiple linked workspaces
The tree structure keeps track of the decisions made by the analyst
Interaction tasks can be performed either in each node or in the whole tree
The system interaction promotes the creation of classification trees that help to interpret the visualization in a partitioned manner
By promoting multiple views exploration, the systems allows
scalability and flexibility.
(SG-05)
15/30
• Introduction
•Developed System
• Interaction Systematization
• Summarization Features
• Conclusions and Ending
Out
line
Out
line
• Introduction
• Developed System
• Interaction Systematization
• Summarization Features
(SG-05)
16/30Fre
quen
cy p
lot
Fre
quen
cy p
lot
A method that combines selective filtering with automatic frequency calculus within a given selection
The frequency is visually presented through the opacity of the graphical items
Dynamic visual auditingcues can transform the cognition task of view registration into a faster perception
inference task.
(SG-05)
17/30Sta
tist
ics
pres
enta
tion
Sta
tist
ics
pres
enta
tion To summarize the
data, basic statistics are presented over the visualizations
Average (green), standard deviation (yellow), median (cyan) and mode values (magenta)
The use of statistics can characterize an entire visual workspace diminishing cognitive load.
(SG-05)
18/30
X1
X1 = RP1 + MRD
Relevance = 0
X0 = RP0
Relevance = 1
X0
X2
X3
Null RP2 Not Considered
Dist = 1
Relevance = - 1
The relevance point is over the attribute value
The distance is equal the Maximum relevance
distance The distance is the maximum possibleRelevance = 1 + 0 + (-1) =
0/3 = 0
The data is presented according to its relevance to a user’s defined set of interesting points
Rel
evan
ce p
lot
Rel
evan
ce p
lot
The Relevance Plot can be used to determine speculative queries in sets where
categorical selections are not efficient.
(SG-05)
19/30
• Introduction
•Developed System
• Interaction Systematization
• Summarization Features
• Conclusions and Ending
Out
line
Out
line
• Introduction
• Developed System
• Interaction Systematization
• Conclusions and Ending
(SG-05)
20/30
Con
clus
ions
Con
clus
ions
The Visualization Tree systematization can help to diminish InfoVis techniques limitations
The interaction, exploration and summarization functionalities, together, can be considered a step further in multivariate visual analysis
The future of InfoVis do not rely on revolutionary new techniques but on integrated systematizations presenting interaction and summarization capabilities
(SG-05)
21/30
The
End
The
End
Thanks for coming
All this information did not fit in the paper, so the tool (MSWindows) can be downloaded at
http://www.gbdi.icmc.usp.br/~junio/vistree
or at
http://vistree.got.to (alias) We have validated most of the system’s
features based on literature knowledge. However…
To do: perform a systematic evaluation of the tool usability