F inding patterns in temporal data

1. Finding patterns in temporal data 27th HCIL Symposium May 27, 2010 Krist wongsuphasawat Taowei david wang Catherine plaisant Ben shneiderman Human-computer interaction lab University of maryland

2. Finding patterns in temporal data 27th HCIL Symposium May 27, 2010 Krist wongsuphasawat Taowei david wang Catherine plaisant Ben shneiderman Human-computer interaction lab University of maryland

3. Temporal categorical data Category • A type of time series Numerical Event Patient ID: 45851737 Stock: Microsoft Event 04/26/2010 10:00 31.03 04/26/2010 10:15 31.01 04/26/2010 10:30 31.02 04/26/2010 10:45 31.08 04/26/2010 11:00 31.16 12/02/2008 14:26 Arrival 12/02/2008 14:36 Emergency 12/02/2008 22:44 ICU 12/05/2008 05:07 Floor 12/14/2008 06:19 Exit Time Arrival Emergency ICU Floor Exit

4. Temporal categorical data Electronic Health Records: symptoms, treatment, lab test Traffic incident logs: arrival/departure time of each unit Student records: course, paper, proposal, defense, etc. Others: web logs, usability study logs, etc.

5. 10+ years work on temporal visualization (mostly on Electronic Health Records)

6. Lifelines Single Record [Plaisant et al. 1998] http://www.cs.umd.edu/hcil/lifelines

7. LifeLines – Single Patient

8. working with physicians at Washington Hospital Center

9. Example data • Patient transfers

10. tasks • Example: Finding “Bounce backs” ICU Floor ICU within 2 days

11. Lifelines 2 record record record record record [Wang et al. 2008, 2009] http://www.cs.umd.edu/hcil/lifelines2

12. ARF (Align-Rank-Filter) Framework Multiple Records Temporal Summary LifeLines2 – Search and Visualize

13. Alignment • Sentinel events as reference points Time August June July Patient #45851737 Arrival Emergency ICU Floor Exit Patient #43244997 Arrival Emergency ICU Floor Exit

14. Alignment (2) • Time shifting Time 2 M 0 1 M Patient #45851737 Admit Emergency ICU Floor Exit Patient #43244997 Admit Emergency ICU Floor Exit

15. similan record record record record record [Wongsuphasawat & Shneiderman 2009] http://www.cs.umd.edu/hcil/similan

16. Similan – Search by Similarity

17. Similan – Search by Similarity

18. Finding “bounce backs” Before After • Much faster to specify new query • Visualizing the results gives better understanding

19. User studies: search LifeLines2 Similan Exact Similarity-based MUST have A, B, C SHOULD have A, B, C Query Query Record#2 Record#2 Record#1 moresimilar Record#1 Record#3 Record#3

20. User studies: search LifeLines2 Similan Exact Similarity-based MUST have A, B, C SHOULD have A, B, C Query Query Record#2 1 Record#2 Record#1 moresimilar Record#1 Record#3 Record#3

21. New stuff Needs for an overview -> LifeFlow!

22. tasks • Example: Finding “Bounce backs” • Other questions Arrival ? ICU Floor ICU ICU ? ? within 2 days

23. Lifeflow record visualize record record Display the aggregation record record record record record Aggregate Merge multiple records into tree

24. Aggregate • Aggregate by prefix #1 #2 #3 #4 Example with 4 records

25. Aggregate • Aggregate by prefix #1 #2 #3 #4

26. Visualize • Inspired by the Icicle tree [Fekete 2004] Number of files

27. Visualize (2) • Use horizontal axis to represent time • Video

28. Demo – lifeflow When the lines are combined into flow

29. Future work ICU Intermediate ICU Intermediate • Comparison Floor Jan-Mar 2008 April-June 2008

30. Take-away message Information visualization is a powerful way to explore temporal patterns. You can work with us on new case studies.

31. Temporal categorical data Electronic Health Records: symptoms, treatment, lab test Traffic incident logs: arrival/departure time of each unit Student records: course, paper, proposal, defense, etc. Others: web logs, usability study logs, etc.

32. Example – traffic incidents

33. acknowledgement Dr. Phuong Ho, Dr. Mark Smith, David Roseman Washington Hospital Center http://www.whcenter.org National Institutes of Health (NIH) - Grant CA147489 http://www.nih.gov Michael Pack, Michael VanDaniker Center for Advanced Tranportation Technology Lab (CATT Lab) http://www.cattlab.umd.edu

34. Take-away message Information visualization is a powerful way to explore temporal patterns. You can work with us on new case studies. • More demos this afternoon • {kristw, tw7, plaisant, ben}@cs.umd.edu • http://www.cs.umd.edu/hcil/temporalviz

35. Q&A Questions? • {kristw, tw7, plaisant, ben}@cs.umd.edu http://www.cs.umd.edu/hcil/temporalviz

36. Thank you Thank you

37. Backup slides Junkyard...

38. Lifelines2 • 8 case studies • Bounce backs • Step ups • BIPAP • Etc.

39. Dr. P

40. LifeLines2’s Temporal Summary [Wang et al. 2009] Continuum’s Histogram [Andre 2007] Does not help exploring sequential patterns Needs a new overview

41. User studies • 8 Extensive case studies • Compared LifeLines2 with Similan • Learn advantages & disadvantages • Drawing is preferred • Clear cut off points is needed • Working on improvements • Flexible temporal search

42. similan • Compared with LifeLines2 in an experiment • Learn advantages & disadvantages • Drawing is preferred • No clear cut off points • Working on improvements • Flexible temporal search

43. Lifeflow Aggregate Merge multiple records into tree Visualize Display the tree

44. approaches Exact Search Similarity-based Search MUST have A, B, C SHOULD have A, B, C Query Query Record#1 Record#2 moresimilar Record#2 Record#1 Record#3 Record#3

45. Research Question#1 Prelim. + proposed work motivation conclusion Research questions Research Question#2 Prelim. + proposed work

47. Expected Contributions Design of visual representations, user interfaces and interaction techniques Algorithms for flexible temporal search Evaluation results Open new directions for exploring temporal categorical data

48. Needs for an overview • We learn

49. Needs Visualize overview or show summary Where should I start?

50. Temporal visualizations Background and related work