The document discusses visual analytics and information visualization, describing how new tools can provide insights from data through the combination of visualization, human interaction, and data mining. It outlines challenges in creating meaningful visual displays of massive data, enabling interaction, and developing process models for discovery. The field of visual analytics uses visual representations and interaction to help analysts gain insights into datasets and solve analytical problems.
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
Visual Analytics: New Tools for Gaining Insight from Your Data
1. Visual Analytics:
New Tools for Gaining Insight from Your Data
Ben Shneiderman ben@cs.umd.edu
Founding Director (1983-2000), Human-Computer Interaction Lab
Professor, Department of Computer Science
Member, Institute for Advanced Computer Studies
University of Maryland
College Park, MD 20742
2. Visual Analytics:
New Tools for Gaining Insight from Your Data
Ben Shneiderman ben@cs.umd.edu
Twitter: @benbendc
University of Maryland
College Park, MD 20742
7. Information Visualization
• Visual bandwidth is enormous
• Human perceptual skills are remarkable
• Trend, cluster, gap, outlier...
• Color, size, shape, proximity...
• Three challenges
• Meaningful visual displays of massive data
• Interaction: widgets & window coordination
• Process models for discovery
8. Information Visualization & Visual Analytics
• Visual bands
• Human percle
• Trend, clus..
• Color, size,..
• Three challe
• Meaningful vi
• Interaction: w
• Process mo
1999
9. Information Visualization & Visual Analytics
• Visual bandwidth is enormous
• Human perceptual skills are remarkable
• Trend, cluster, gap, outlier...
• Color, size, shape, proximity...
• Three challenges
• Meaningful visual displays of massive da
• Interaction: widgets & window coordinati
• Process models for discovery
1999 2004
10. Information Visualization & Visual Analytics
• Visual bandwidth is enormous
• Human perceptual skills are remarkable
• Trend, cluster, gap, outlier...
• Color, size, shape, proximity...
• Three challenges
• Meaningful visual displays of massive data
• Interaction: widgets & window coordination
• Process models for discovery
1999 2004 2010
11. Business takes action
• General Dynamics buys MayaViz
• Agilent buys GeneSpring
• Google buys Gapminder
• Oracle buys Hyperion
• Microsoft buys Proclarity
• InfoBuilders buys Advizor Solutions
• SAP buys (Business Objects buys
Xcelsius & Inxight & Crystal Reports )
• IBM buys (Cognos buys Celequest) & ILOG
• TIBCO buys Spotfire
19. Information Visualization: Data Types
• 1-D Linear
.
Document Lens, SeeSoft, Info Mural
• 2-D Map GIS, ArcView, PageMaker, Medical imagery
• 3-D World CAD, Medical, Molecules, Architecture
zi Vc S
i
• Multi-Var Spotfire, Tableau, GGobi, TableLens, ParCoords,
• Temporal LifeLines, TimeSearcher, Palantir, DataMontage
• Tree Cone/Cam/Hyperbolic, SpaceTree, Treemap
• Network Pajek, JUNG, UCINet, SocialAction, NodeXL
zi V f nI
o
infosthetics.com flowingdata.com infovis.org
www.infovis.net/index.php?lang=2
24. Temporal Data: TimeSearcher 1.3
• Time series
• Stocks
• Weather
• Genes
• User-specified
patterns
• Rapid search
25. Temporal Data: TimeSearcher 2.0
• Long Time series (>10,000 time points)
• Multiple variables
• Controlled precision in match
(Linear, offset, noise, amplitude)
29. LifeFlow: Aggregation Strategy
Temporal
Categorical Data
(4 records)
LifeLines2 format
Tree of Event
Sequences
LifeFlow Aggregation
www.cs.umd.edu/hcil/lifeflow
44. Treemap: WHC Emergency Room
(6304 patients in Jan2006)
Group by Admissions/MF, size by service time, color by age
45. Treemap: WHC Emergency Room
(6304 patients in Jan2006) (only those service time >12 hours)
Group by Admissions/MF, size by service time, color by age
53. Discovery Process: Systematic Yet Flexible
Preparation
• Own the problem & define the schedule
• Data cleaning & conditioning
• Handle missing & uncertain data
• Extract subsets & link to related information
54. SocialAction
• Integrates statistics
& visualization
• 4 case studies, 4-8 weeks
(journalist, bibliometrician, terrorist analyst,
organizational analyst)
• Identified desired features, gave strong positive
feedback about benefits of integration
www.cs.umd.edu/hcil/socialaction
Perer & Shneiderman, CHI2008, IEEE CG&A 2009
68. No Location Philadelphia
Innovation Clusters: People, Locations, Companies
11,000 nodes
26,000 links
Pharmaceutical/Medical
Pittsburgh Metro
Westinghouse Electric
69. No Location Philadelphia
Innovation Clusters: People, Locations, Companies
Pharmaceutical/Medical
Pittsburgh Metro
Westinghouse Electric
70. No Location Philadelphia
Innovation Clusters: People, Locations, Companies
Patent
Tech
Navy SBIR (federal)
PA DCED (state)
Related patent
2: Federal agency
Pharmaceutical/Medical 3: Enterprise
Pittsburgh Metro 5: Inventors
9: Universities
10: PA DCED
11/12: Phil/Pitt metro cnty
13-15: Semi-rural/rural cnty
17: Foreign countries
19: Other states
Westinghouse Electric
75. Analyzing Social Media Networks with NodeXL
I. Getting Started with Analyzing Social Media Networks
1. Introduction to Social Media and Social Networks
2. Social media: New Technologies of Collaboration
3. Social Network Analysis
II. NodeXL Tutorial: Learning by Doing
4. Layout, Visual Design & Labeling
5. Calculating & Visualizing Network Metrics
6. Preparing Data & Filtering
7. Clustering &Grouping
III Social Media Network Analysis Case Studies
8. Email
9. Threaded Networks
10. Twitter
11. Facebook
12. WWW
13. Flickr
14. YouTube
15. Wiki Networks
www.elsevier.com/wps/find/bookdescription.cws_home/723354/description
76. Social Media Research Foundation
Researchers who want to
- create open tools
- generate & host open data
- support open scholarship
Map, measure & understand
social media
Support tool projects to
collection, analyze & visualize
social media data.
smrfoundation.org
79. Discovery Process: Systematic Yet Flexible
Preparation
• Own the problem & define the schedule
• Data cleaning & conditioning
• Handle missing & uncertain data
• Extract subsets & link to related information
80. Discovery Process: Systematic Yet Flexible
Preparation
• Own the problem & define the schedule
• Data cleaning & conditioning
• Handle missing & uncertain data
• Extract subsets & link to related information
Purposeful exploration – Hypothesis testing
• Range & distribution
• Relationships & correlations
• Clusters & gaps
• Outliers & anomalies
• Aggregation & summary
• Split & trellis
• Temporal comparisons & multiple views
• Statistics & forecasts
81. Discovery Process: Systematic Yet Flexible
Preparation
• Own the problem & define the schedule
• Data cleaning & conditioning
• Handle missing & uncertain data
• Extract subsets & link to related information
Purposeful exploration – Hypothesis testing
• Range & distribution
• Relationships & correlations
• Clusters & gaps
• Outliers & anomalies
• Aggregation & summary
• Split & trellis
• Temporal comparisons & multiple views
• Statistics & forecasts
Situated decision making - Social context
• Annotation & marking
• Collaboration & coordination
• Decisions & presentations
82. UN Millennium Development Goals
To be achieved by 2015
• Eradicate extreme poverty and hunger
• Achieve universal primary education
• Promote gender equality and empower women
• Reduce child mortality
• Improve maternal health
• Combat HIV/AIDS, malaria and other diseases
• Ensure environmental sustainability
• Develop a global partnership for development
84. For More Information
• Visit the HCIL website for 650 papers & info on videos
www.cs.umd.edu/hcil
• Conferences & resources: www.infovis.org
• See Chapter 14 on Info Visualization
Shneiderman, B. and Plaisant, C., Designing the User Interface:
Strategies for Effective Human-Computer Interaction:
Fifth Edition (2010) www.awl.com/DTUI
• Edited Collections:
Card, S., Mackinlay, J., and Shneiderman, B. (1999)
Readings in Information Visualization: Using Vision to Think
Bederson, B. and Shneiderman, B. (2003)
The Craft of Information Visualization: Readings and Reflections
"The IN Cell Analyzer automated microscope was used to identify proteins influencing the division of human cells. After the images were analyzed, quantitative results were transferred to Spotfire DecisionSite. This screen revealed the previously unknown involvement of the retinol binding protein RBP1 in cell cycle control.(Stubbs S, & Thomas N. 2006 Methods in Enzymology; 414:1-21.) Retinol a form of Vitamin A plays a crucial role in vision and during embryonic development"
Contrast and Creatinine dataset In some diagnostic radiology procedures, patients are injected contrast material. However, some patients develop adverse side effects to the contrast material. One serious side effect is renal failure, which is detected by high creatinine levels in a patient's blood. This adverse effect usually occur within two weeks after the radiology contrast. WHC is interested in finding the proportion of patients who exhibit this condition in historical records. Screenshots 1-aligned-ranked.png: We align by the 1st occurrence of radiology contrast and rank by the number of creatinine high (CREAT-H) events to bring the most severe patients to the top. We realize two things: (1) some patients have more than 1 "Radiology Contrast" events, and (2), some patients have consistently high creatinine readings (chronic kidney failure). 2-aligned(all)-distribution-selected.png We align by all occurrences of raiology contrast, and then show the temporal summary of CREAT-H events. The patients are presented in 4 exclusive sets in the summary: those who have CREAT-H only before alignment, only after alignment, both before and after, and neither. We then select from the "only after" summary the patients who have at least one CREAT-H event within 2 weeks of any "Radiology Contrast" event. There are 421 patients.
Using LifeFlow, 7,041 patients are aggregated into this visualization and LifeFlow immediately reveal the most common pattern, which you could not do easily in SQL. You could easily notice this huge pattern “Arrival -> ER -> Exit”, meaning patients who visited with minor injuries or simple conditions and left the hospital immediately after receiving their treatment. When hovering the mouse over, LifeFlow displays a tooltip that gives more information, such as number of patients and other statistics, and also shows the distribution of the patients. As the horizontal gap represents time, you can see from the distribution that some patients left the hospital very quickly after visiting the emergency room while some of them stayed longer. *optional The second most common pattern is “Arrival (Blue) -> ER (Pink) -> Floor (Green) -> Exit (Cyan)”, meaning patients who were admitted to observe the conditions and then everything went well so they left the hospital. You can also use the horizontal gap to compare these patients with the patients who exit from the emergency room. Comparing the gap from pink to cyan and pink to green, you can see that the gap from pink to green is smaller than pink to cyan, so the patients were transferred to Floor faster than exit the hospital in average. You have seen the two most common cases, now I will remove the common patterns so we can analyze the less frequent patterns.
After removing all the common cases, we have 344 patients left. These are mostly the patients who were admitted. There are many information that I can explain from this visualization here, but I will go straight into the case that our physician partners are mostly interested in. The mouse is pointing at this sequence, which represents the “bounce backs” patients, meaning patients who were transferred from ICU to Floor because they seemed to get better, however, they were transferred back to the ICU. So the physician are interested in finding these patients to analyze what made them made the wrong decisions. *optional Another case is the step ups, which means the patients whose level of care were escalated to higher level, you can see from the visualization that there were patients who were transferred from ER to Floor (green) to ICU (red) and IMC (orange). The number of these patients and the average transferred time could be compare to the hospital standards to measure the quality of care.
Ben: This slide is optional. You can use it to show that when you click on the bounce backs patients, you can get the details of each patient in LifeLines2 view.
Another interesting feature is you can align by a particular event. For example, if you want to know what happened before and after the patients went to the ICU, you can align by ICU. The dash line separate between what happened before and what happened after. You can see that the ICU patients mostly came from the ER (pink), and most of them were transferred to Floor (green) after that. Unfortunately, some of them died after they were transferred to the ICU (black). From this visualization, you may notice a small pattern in the bottom. Let me zoom in.
So this patient was dead before transferred to the ICU, which is impossible. Of course, this must be problem with data entry. But we may never notice it if the data are hidden in the database. Therefore, you can see that LifeFlow support this kind of analysis by giving overview, showing common trends, providing summary of every sequences, you can do SQL and calculate average for every transfer if you like, but in LifeFlow, it is right there, you just need to move your mouse over. showing every possible transfer pattern and may led you to a discovery of surprising pattern.
Live Demonstration
Aligning sales and marketing is essential for success. The graph on the left shows sales people linked to opportunities, including industry. The thicker the line, the higher the probability of closing the deal. The larger the dollar sign, the bigger the deal. Sullivan, Vazquez and Distefano are performing the best. The upper right shows the number of deals by stage in the sales cycle. The blue bubble chart shows potential revenue by marketing program and stage in the sales cycle. Search engine optimization and inbound links from Web sites have the biggest impact. Armed with this information, marketing managers can advertise to the financial services and manufacturing sectors through specific tactics, and sales managers can see the performance of the reps and the industries where they are successful.
Chapter 3, Figure 1 (page 6). A NodeXL social media network diagram of relationships among Twitter users mentioning the hashtag “#WIN09” used by attendees of a conference on Network Science at NYU in September 2009. Each user’s node is sized proportional to the number of tweets they have ever made to that date.
Figure 1. (a) Harel-Koren (HK) fast multi-scale layout of a clustered network of Twitter users, using color to differentiate among the vertices in different clusters. The layout produces a visualization with overlapping cluster positions. . (b) Group-in-a-Box (GIB) layout of the same Twitter network: clusters are distributed in a treemap structure that partitions the drawing canvas based on the size of the clusters and the properties of the rendered layout. Inside each box, clusters are rendered with the HK layout.
Figure 3. The 2007 U.S. Senate co-voting network graph, visualized with the GIB layout. The group in each box represents senators from a given U.S. region (1: South; 2: Midwest; 3: Northeast; 4: Mountain; 5: Pacific) and individual groups are displayed using the FR layout. Vertices colors represent the senators’ party affiliations (blue: Democrats; red: Republicans; orange: Independent) and their size is proportional to betweenness centrality. Edges represent percentage of agreement between senators: (a) above 50%; (b) above 90%..
Figure 13.20. NodeXL cluster visualization showing three Flickr tag clusters, each representing a different context for “mouse”. Figure 13.21. NodeXL display of Isolated clusters for three different contexts for the “mouse” tag in Flickr: mouse animal, computer mouse, and Mickey Mouse Disney character.
Chapter 3, Figure 1 (page 6). A NodeXL social media network diagram of relationships among Twitter users mentioning the hashtag “#WIN09” used by attendees of a conference on Network Science at NYU in September 2009. Each user’s node is sized proportional to the number of tweets they have ever made to that date.