outflow: exploring flow, factors and outcome of temporal event sequences

72
m Outflow Krist Wongsuphasawat HCIL, University of Maryland David Gotz IBM Research Exploring Flow, Factors and Outcomes of Temporal Event Sequences InfoVis 2012 Seattle, WA

Upload: krist-wongsuphasawat

Post on 27-Jan-2015

117 views

Category:

Technology


6 download

DESCRIPTION

My presentation at IEEE VisWeek 2012 in Seattle, WA //// Abstract: Event sequence data is common in many domains, ranging from electronic medical records (EMRs) to sports events. Moreover, such sequences often result in measurable outcomes (e.g., life or death, win or loss). Collections of event sequences can be aggregated together to form event progression pathways. These pathways can then be connected with outcomes to model how alternative chains of events may lead to different results. This paper describes the Outflow visualization technique, designed to (1) aggregate multiple event sequences, (2) display the aggregate pathways through different event states with timing and cardinality, (3) summarize the pathways’ corresponding outcomes, and (4) allow users to explore external factors that correlate with specific pathway state transitions. Results from a user study with twelve participants show that users were able to learn how to use Outflow easily with limited training and perform a range of tasks both accurately and rapidly.

TRANSCRIPT

Page 1: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow

Krist Wongsuphasawat HCIL, University of Maryland

David Gotz IBM Research

Exploring Flow, Factors and Outcomes of Temporal Event Sequences

InfoVis 2012 Seattle, WA

Page 2: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Events

Page 3: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Event | 12:15 p.m. Lunch

Page 4: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Event Sequences Event Event Event

Page 5: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Daily Activity

7:30 a.m. Wake Up

7:45 a.m. Exercise

8:15 a.m. Go to work

Page 6: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Soccer Game

90th minute Team A scores

25th minute Team B scores

10th minute Team A scores

Page 7: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Game #1

Time

10th minute Goal

90th minute Goal

25th minute Concede

Soccer Game

Page 8: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Goal

Game #1

Concede Goal

Goal

Game #2

Goal Concede

Time

Goal

Game #3

Concede Concede

Concede

Game #n

Goal Goal Goal

Many games

Page 9: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

with outcome

Game #1

Game #2

Time

Game #3

Game #n

Lose (0)

Win (1)

Win (1)

Win (1)

Goal Concede Goal

Goal Goal Concede

Goal Concede Concede

Concede Goal Goal Goal

Page 10: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

7 event types

823543 combinations

7 events per entity

Page 11: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Enjoy!

Page 12: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

consumable

Page 13: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Overview / Summary

Event Sequences with Outcome

Page 14: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

7 Steps

Page 15: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 1 | Aggregation

Page 16: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Entity #1

Entity #2

Entity #4

Entity #3

Entity #5

Entity #6

Entity #n

Entity #7

Outflow Graph

Event Sequences

Page 17: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Assumption •  Events are persistent.

e1

Entity #1

e2 e3

Entity #1

Page 18: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Assumption •  Events are persistent.

e1

Entity #1

e2 e3

e1

Entity #1

e1 e1

Page 19: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Assumption •  Events are persistent.

e1

Entity #1

e2 e3

e1

Entity #1

e1 e2

e1 e2

Page 20: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Assumption •  Events are persistent.

e1

Entity #1

e2 e3

e1

Entity #1

e1 e2

e1 e2 e3

Page 21: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Assumption •  Events are persistent.

e1

Entity #1

e2 e3

e1

Entity #1

e1 e2

e1 e2 e3

[e1]

[e1, e2]

[e1, e2, e3] States

Page 22: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Select alignment point Pick a state

What are the paths that led to ?

What are the paths after ?

Soccer: Goal, Concede, Goal

Example

Page 23: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Select alignment point Pick a state

What are the paths that led to ?

What are the paths after ?

or just an empty state []

Page 24: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e1, e2, e3]

Alignment Point

Page 25: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e1, e2, e3]

[e1, e2]

[e1, e2, e3, e5]

[e1]

[ ]

Alignment Point

1 entity

Page 26: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e1, e3]

Alignment Point

2 entities

[e1, e2, e3]

[e1, e2]

[e1, e2, e3, e5]

[e1]

[ ]

Page 27: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e1, e2, e3, e4]

Alignment Point

[e3]

3 entities

[e1, e3]

[e1, e2, e3]

[e1, e2]

[e1, e2, e3, e5]

[e1]

[ ]

Page 28: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e2, e3]

[e2]

Alignment Point

n entities

[e1, e2, e3, e4]

[e3]

[e1, e3]

[e1, e2, e3]

[e1, e2]

[e1, e2, e3, e5]

[e1]

[ ]

Page 29: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Outflow Graph

[e2, e3]

[e2]

Alignment Point

n entities

Average outcome Average time Number of entities

= 0.4 = 10 days = 10

[e1, e2, e3, e4]

[e3]

[e1, e3]

[e1, e2, e3]

[e1, e2]

[e1, e2, e3, e5]

[e1]

[ ]

layer

Page 30: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Soccer Results

2-1

2-0

1-1

0-2

2-2

3-1

1-0

0-1

0-0

Alignment Point

Page 31: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 2 | Visual Encoding

Page 32: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Alignment Future Past

e1!e2!

e1!

e2!

e1!e2!e3!

e1!e2!e4!

Color is outcome measure.

Node’s height is number of entities.

Time edge’s width is duration of transition.

Node’s horizontal position shows sequence of states.

time edge

link edge

End of path

Page 33: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 3 | Graph Drawing

Page 34: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 35: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 36: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

3.1 Sugiyama’s heuristics •  Directed Acyclic Graph (DAG) layout

–  Sugiyama, K., Tagawa, S. & Toda, M., 1981. Methods for Visual Understanding of Hierarchical System Structures. IEEE Transactions on Systems, Man, and Cybernetics, 11(2), p.109-125.

•  Reduce edge crossing

Page 37: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

41 crossings

Page 38: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

12 crossings

Page 39: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 40: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

3.2 Force-directed layout •  Spring simulation

x

Each node is particle.

Total force = Force from edges - Repulsion between nodes

Page 41: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 42: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 43: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

3.3 Edge Routing •  Avoid unnecessary crossings

Reroute

Page 44: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

3.3 Edge Routing •  After routing

Page 45: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 46: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 47: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 4 | Interactions

Page 48: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Interactions •  Panning •  Zooming •  Brushing •  Pinning •  Tooltip •  Event type selection

Page 49: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Demo

Page 50: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 5 | Simplification

Page 51: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Node Clustering •  Cluster nodes in each layer •  Similarity measure: Outcome, etc. •  Threshold (0-1)

Page 52: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 53: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 54: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 6 | Factors

Page 55: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Entity #1

Factors Time

[e1] [e1, e2] [e1, e2, e3]

Factor 1 Factor 2 Factor 3 Factor 4

Page 56: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Patient #1

Factors Time

Which factors are correlated to each state?

Yellow Injury Red Substitution

[e1] [e1, e2] [e1, e2, e3]

Page 57: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Which keywords are correlated to each document?

Information Retrieval

State 1 … …

State 2 … … …

State 3 … … …

Doc#1 Doc#2 Doc#3

Factor xxx

Which factors are correlated to each state?

Page 58: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Present factors

[e1,e2,e3]

[e1,e2]

[e1,e3]

[e2,e3]

[e1,e2,e3,e4]

[e1,e2,e3,e5]

[e1]

[e2]

[e3]

[ ]

Alignment Point

Factor 1

Page 59: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Absent factors

[ ]

Alignment Point

Factor 2

Factor 2 [e1,e2,e3]

[e1,e2]

[e1,e3]

[e2,e3]

[e1,e2,e3,e4]

[e1,e2,e3,e5]

[e1]

[e2]

[e3]

Page 60: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

tf-idf •  Term frequency

Number of times a term t appear in the document

Number of terms in the document

Number of documents

Number of documents that has the term t + 1 log ( )

tf

idf

=

=

•  Inverse document frequency

Page 61: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Score based on tf-idf •  Ratio (presence)

Number of entities with factor f before state

Number or entities in the state

Number of states

Number of states preceded by factor f + 1 log ( )

Rp

R-1

=

=

•  Inverse state ratio (presence)

sp

Page 62: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Page 63: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Step 7 | User Study

Page 64: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

User Study •  Goal:

Evaluate Outflow’s ability to support event sequence analysis tasks

•  12 participants •  60 minutes each •  9 tasks + 7 training tasks •  Questionnaire

Page 65: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Results •  Accurate:

3 mistakes from 108 tasks

•  Fast: Average 5-60 seconds

•  Findings: –  From video –  Different outcomes for each incoming paths –  Etc.

Page 66: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Future Work •  Integration with prediction algorithm •  Additional layout techniques •  Advanced factor analysis •  Deeper evaluations with domain experts

Page 67: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Conclusions •  Event sequences with outcome •  Outflow

–  Interactive visual summary –  Explore flow & outcome –  Factors –  Multi-step layout process

•  Not specific to sports

Contact: @kristwongz [email protected] [email protected]

Page 68: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Patient #1

Time

Aug 1998 Ankle Edema

Jan 1999 Weight Loss

Oct 1998 Cardiomegaly

Heart failure (CHF) patient Die (0)

Page 69: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Event Sequences

and more…

Medical Transportation

Education

Web logs

Sports

Logistics

Page 70: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Acknowledgement •  Charalambos (Harry) Stavropoulos •  Robert Sorrentino •  Jimeng Sun •  Comments from HCIL colleagues

Page 71: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

Conclusions •  Event sequences with outcome •  Outflow

–  Interactive visual summary –  Explore flow & outcome –  Factors –  Multi-step layout process

•  Not specific to medical or sports

Contact: @kristwongz [email protected] [email protected]

Page 72: Outflow: Exploring Flow, Factors and Outcome of Temporal Event Sequences

m

THANK YOU ขอบคุณครับ