• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Visualization for Event Sequences Exploration
 

Visualization for Event Sequences Exploration

on

  • 2,384 views

My talk at the Data Visualization Summit in San Francisco April 11, 2013 ...

My talk at the Data Visualization Summit in San Francisco April 11, 2013
http://theinnovationenterprise.com/summits/data-visualization-sf

----------------
Abstract
----------------
Many aspects of our lives can be captured and described as series of events, or event sequences. These event sequences can be keys to understanding many things: medical services, logistics, sports, user behavior, etc. In this presentation, I will talk about techniques for visualizing event sequences, from simple to advance, and also show examples that demonstrate the power of visualizations in exploring and understanding event sequences.

Statistics

Views

Total Views
2,384
Views on SlideShare
1,638
Embed Views
746

Actions

Likes
7
Downloads
35
Comments
0

5 Embeds 746

http://localhost 383
http://kristw.yellowpigz.com 182
http://kristw-dev.yellowpigz.com 144
https://twitter.com 21
http://kristw-staging.yellowpigz.com 16

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

CC Attribution-NonCommercial LicenseCC Attribution-NonCommercial License

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Visualization for Event Sequences Exploration Visualization for Event Sequences Exploration Presentation Transcript

    • Data Visualization Summit San Francisco, CA Apr 11, 2013 Visualizations forEvent Sequences Exploration Krist Wongsuphasawat Data Visualization Scientist Twitter, Inc. @kristw
    • event% event% event% event% event% event%event% event% Life event% event% event% event% event% event% event% event%
    • Time Event type% ( 7:00 am, Wake up ) event% event% event% event% event% event%event% event% Life event% event% event% event% event% event% event% event%
    • event% event% event% event% event% event%event% event% Life event% event% event% event% event% event% event% event% “Event Sequence”
    • Daily Activity7:30 a.m. 7:45 a.m. 8:30 a.m.Wake Up Exercise Go to work
    • Traffic Incidents9:30 a.m. 9:55 a.m. 10:30 a.m.Notification Units arrived Road cleared
    • http://timeline.national911memorial.org/
    • Event SequencesMedical TransportationSports EducationWeb logs Logistics and more…
    • Outline ? u ences e nt seq hat are ev them? W is ualize Ho w to v a b ig dat Ap ply to
    • Visualization Techniques
    • Event glyphs timelinesequence
    • simple event sequencetimeline.js Horizontal axis = time Glyphs = events http://timeline.verite.co/
    • Event glyphs timeline sequence+ Interval
    • interval •  Car crash (point) time 10 a.m. •  Meeting (interval) 10 – 11 a.m.
    • interval >> widthtraffic incident CATT Lab, University of Maryland -- http://teachamerica.com/VIZ11/VIZ1102Pack/index.htm
    • interval >> widthchronoline.js http://stoicloofah.github.io/chronoline.js/
    • Event glyphs timeline sequence+ Interval width+ Event types
    • types time Nurses’ actions Doctors’ actions They all look similar.
    • types time Nurses’ actions Doctors’ actions Better?
    • The path of protest types >> colorhttp://www.guardian.co.uk/world/interactive/2011/mar/22/middle-east-protest-interactive-timeline
    • types >> colors + shapes http://timeglider.com/widget/timeglider.js
    • Event glyphs timeline sequence+ Interval width+ Event colors shapes types High+ density
    • high density time Too many overlaps and occlusions
    • high density >> facetGoogle Chrome loading scripting rendering & painting Facet Google Chrome > Developer Tools > Timeline
    • high density >> facetLifelines http://www.cs.umd.edu/lifelines
    • high density >> binningBritish History Timeline bin by year
    • high density >> aggregationCloudLines Raw event data Kernel Density Estimation + Importance Func. + Truncation Encode cloud size
    • high density >> aggregationCloudLines (2) Krstajic, M., Bertini, E., & Keim, D. A. (2011). CloudLines: Compact Display of Event Episodes in Multiple Time-Series. IEEE Transactions on Visualization and Computer Graphics, 17(12), 2432.
    • linear Event glyphs timeline sequence non-linear+ Interval width+ Event colors shapes types High+ facet aggregation binning density
    • circular timeline 2008 2009 2010 2011 2012 linear Dec Jan Feb Nov Mar circular Oct Apr repeating patterns Sep May Aug Jun Jul
    • circular timeline (2)Traffic Incidents VanDaniker, M. (2010). Leverage of Spiral Graph for Transportation System Data Visualization. Transportation Research Record: Journal of the Transportation Research Board, 2165, 79–88.
    • stacked timeline 2008 2009 2010 2011 2012 linear 2008 2009 2008 2009 2011 2010 2012 2010 2011 2012
    • stacked timeline (2)Tweet Volume Rios, M., & Lin, J. (2012). Distilling Massive Amounts of Data into Simple Visualizations : Twitter Case Studies. Proceedings of the Workshop on Social Media Visualization (SocMedVis) at ICWSM 2012 (pp. 22–25).
    • linear Event glyphs timeline sequence non-linear+ Interval width+ Event colors shapes types High+ facet aggregation binning density
    • collection 1 2 n Event Event ... Eventsequence sequence sequence
    • collectionmultiple timelines Event sequence #1 Event sequence #2 Event sequence #3 Event sequence #4
    • collection 1 2 n Event Event ... Eventsequence sequence sequence Millions!
    • collection 1 2 n Event Event ... Eventsequence sequence sequence Interactions
    • Interaction #1align
    • Interaction #1align
    • Interaction #1align
    • Interaction #2rank
    • Interaction #2rank Rank by number of events or any criteria
    • Interaction #3filter
    • Interaction #3filter Select only event sequences with events Set your own filters
    • Interaction #4group
    • Interaction #4group 1 2 3 Group by sequence length or any clustering algorithm / properties
    • Interaction #5search •  Simple search ABC –  Sequence matching AABCDEFGH –  Subsequence matching AXAYBZCED •  Regular Expression A B* (C|D)
    • Interaction #5search (2) •  Dynamic X 50% C 75% AB Y 50% D 25%
    • Interaction #5search (2) •  Dynamic X 70% D 50% ABC Y 30% E 50% •  Similarity search Similar to ABCD ABCD ABD ACE …
    • collection 1 2 n Event Event ... Event sequence sequence sequence Interactions Aggregationalign by time rank search filter group
    • aggregation by timetemporal summary Day 1 Day 2 Day 3 Day 4 Day 5 bin & count
    • aggregation by time temporal summaryWang, T. D., Plaisant, C., Shneiderman, B., Spring, N., Roseman, D., Marchand, G., Mukherjee, V., et al. (2009). Temporal Summaries: Supporting Temporal Categorical Searching, Aggregation and Comparison. IEEE Transactions on Visualization and Computer Graphics, 15(6), 1049–1056.
    • collection 1 2 n Event Event ... Event sequence sequence sequence Interactions Aggregationalign by time rank search by sequence filter group
    • aggregation by sequenceLifeFlow e.g. 1) What happened to the patients after they arrived? Arrival! ? ? 2) What happened to the patients before & after ICU? ICU! ? ? ? ?
    • aggregation by sequenceLifeFlow overview / summary Millions of records!
    • Demo LifeFlowWongsuphasawat, K., Guerra Gómez, J. A., Plaisant, C., Wang, T. D., Taieb-Maimon, M., & Shneiderman, B. (2011). LifeFlow: Visualizing an Overview of Event Sequences. Proceedings of CHI2011 (pp. 1747–1756).
    • Demo LifeFlowWongsuphasawat, K., Guerra Gómez, J. A., Plaisant, C., Wang, T. D., Taieb-Maimon, M., & Shneiderman, B. (2011). LifeFlow: Visualizing an Overview of Event Sequences. Proceedings of CHI2011 (pp. 1747–1756).
    • Demo LifeFlowWongsuphasawat, K., Guerra Gómez, J. A., Plaisant, C., Wang, T. D., Taieb-Maimon, M., & Shneiderman, B. (2011). LifeFlow: Visualizing an Overview of Event Sequences. Proceedings of CHI2011 (pp. 1747–1756).
    • aggregation by sequenceLifeFlow profile! home! start! home! photos! home! contact! home!
    • aggregation by sequenceGoogle Analytics profile! start! home! photos! home! contact! http://www.google.com/analytics
    • aggregation by sequenceGoogle Analytics profile! home! start! home! photos! videos! contact! http://www.google.com/analytics
    • aggregation by sequenceGoogle Analytics top pages only height = number of visits http://www.google.com/analytics
    • Event + Outcomesequence
    • Time%Game #1 Win (1) 10th minute 25th minute 90th minute Goal Concede Goal or any sports
    • Time%Game #1 Win (1) Goal% Concede% Goal%Game #2 Win (1) Goal% Goal% Concede%Game #3 Lose (0) Goal% Concede% Concede%Game #n Win (1) Concede% Goal% Goal% Goal%
    • aggregation by sequence with outcomeOutflow (Careflow) overview / summary Event Sequences! with Outcome!
    • Assumption Events are persistent.Record #1 e1% e2% e3%Record #1
    • Assumption Events are persistent.Record #1 e1% e2% e3%Record #1 e1% e1% e1%
    • Assumption Events are persistent.Record #1 e1% e2% e3%Record #1 e1% e1% e1% e2% e2%
    • Assumption Events are persistent.Record #1 e1% e2% e3%Record #1 e1% e1% e1% e2% e2% e3%
    • Assumption Events are persistent.Record #1 e1% e2% e3%Record #1 e1% e1% e1% [e1] e2% e2% e3%States [e1, e2] [e1, e2, e3]
    • Select alignment point Pick a stateWhat are the paths What are the pathsthat led to ? after ? Example Soccer: Goal, Concede, Goal
    • Outflow Graph Alignment Point [e1, e2, e3]!
    • 1%record% Outflow Graph Alignment Point [e1]! [e1, e2]![ ]! [e1, e2, e3]! [e1, e2, e3, e5]!
    • 2%records% Outflow Graph Alignment Point [e1]! [e1, e2]![ ]! [e1, e3]! [e1, e2, e3]! [e1, e2, e3, e5]!
    • 3%records% Outflow Graph Alignment Point [e1]! [e1, e2]! [e1, e2, e3, e4]![ ]! [e1, e3]! [e1, e2, e3]! [e1, e2, e3, e5]! [e3]!
    • n%records% Outflow Graph Alignment Point [e1]! [e1, e2]! [e1, e2, e3, e4]![ ]! [e2]! [e1, e3]! [e1, e2, e3]! [e1, e2, e3, e5]! [e3]! [e2, e3]!
    • n%records% Outflow Graph Alignment Point [e1]! [e1, e2]! [e1, e2, e3, e4]![ ]! [e2]! [e1, e3]! [e1, e2, e3]! [e1, e2, e3, e5]! [e3]! [e2, e3]! Average outcome = 0.4 Average time = 10 days No. of records = 10
    • Soccer Results Alignment Point 1-0! 2-0! 2-2!0-0! 1-1! 2-1! 3-1! 0-1! 0-2!
    • Past& Future& Alignment% Node’s horizontal position shows sequence of states.% e1! e2! e3! End of path%e1! e1! e2! 7me% link% e1! Node’s height is edge% edge% e2! number of records.% e4!e2! Color is outcome Time edge’s width is measure.% duration of transition.%
    • Wongsuphasawat, K., & Gotz, D. (2012).Exploring Flow, Factors, and Outcomes of Temporal Event Sequences with the Outflow Visualization. IEEE Transactions on Visualization and Computer Graphics, 18(12), 2659–2668.
    • collection 1 2 n Event Event ... Event sequence sequence sequence Interactions Aggregationalign by time rank search by sequence filter group + Outcome
    • Application toBig Data Analysis
    • Something sounds simple X magnitude of big data = Big mess & Big reward
    • Event Sequence Analysis ateBay CheckoutProcStep1 PaymentReview CheckoutProcStep2 CheckoutProcStep3 PaymentConfirm CheckoutProcStep4 CheckoutProcStep5 CheckoutProcStep6 CheckoutSuccess
    • eBay Event Sequence Analysis at alignment Shen, Z., Wei, J., Sundaresan, N., & Ma, K.-L. (2012). Visual analysis of massive web session data.IEEE Symposium on Large Data Analysis and Visualization (LDAV), 65–72.
    • Event Sequence Analysis atTwitter •  Data –  TBs of session logs everyday •  Complexity –  millions of sessions per day –  1000+ types of events –  long sessions •  Goal –  Overview of how users are using Twitter •  Technique –  LifeFlow Simplify!
    • Event Sequence Analysis atTwitter (2) •  So far –  millions of sessions per day –  millions of sessions on the same screen –  1000+ types of events –  simplified sets of events •  e.g., pages only, selected pages only –  long sessions –  limited session length to 10-20 events
    • Event Sequence Analysis atTwitter (3) Session%Start% Page%A% Page%B% Page%C% Page%B% Page%A% Page%D% Page%C% Page%D% Page%B% Page%C% Page%D% Page%C% *fake data
    • Event Sequence Analysis atTwitter (4) •  Implementation –  Hadoop  –  Web-based (js) •  More –  Stored preprocessed data in smaller db (MySQL/Vertica) Interactive MySQL / HDFS Vertica Visualization Batch pig scripts
    • Takeaway Messages•  Life is full of event sequences.•  How to visualize an event sequence Krist Wongsuphasawat krist.wongz@gmail.com @kristw
    • linear Event glyphs timeline sequence non-linear+ Interval width+ Event colors shapes types High+ facet aggregation binning density
    • Takeaway Messages•  Life is full of event sequences.•  How to visualize an event sequence•  How to visualize collection of event seq. Krist Wongsuphasawat krist.wongz@gmail.com @kristw
    • collection 1 2 n Event Event ... Event sequence sequence sequence Interactions Aggregationalign by time rank search by sequence filter group + Outcome
    • Takeaway Messages•  Life is full of event sequences.•  How to visualize an event sequence•  How to visualize collection of event seq.•  Applicable to big data•  New techniques happen everyday. Krist Wongsuphasawat krist.wongz@gmail.com @kristw
    • Smurf Communism - Wikipediadelete keep … http://notabilia.net/
    • http://www.evolutionoftheweb.com
    • Takeaway Messages•  Life is full of event sequences.•  How to visualize an event sequence•  How to visualize collection of event seq.•  Applicable to big data•  New techniques happen everyday. Krist Wongsuphasawat krist.wongz@gmail.com @kristw