Upload
ngokiet
View
232
Download
5
Embed Size (px)
Citation preview
INFOVIS8803DV > SPRING 17
TEMPORAL & SEQUENTIAL
DATA VISUALIZATION
Prof. Rahul C. Basole
CS/MGT 8803-DV > February 1, 2017
INFOVIS8803DV > SPRING 17
What is Time Series Data?
• There is a chronological component to the data set.
• Examples?
75 % of 4,000 samples of
graphics from newspapers and
magazines were time-series data!
(1974-1980)
Sunspot activity
GT-UGA outcomes
Stock pricesEvents!
INFOVIS8803DV > SPRING 17
Timeviz.net Browser
http://timeviz.net/
INFOVIS8803DV > SPRING 17
Standard Presentation
Mom and Apple Pie!
time
mystery
variable
INFOVIS8803DV > SPRING 17
Lots of Examples
INFOVIS8803DV > SPRING 17
A Taxonomy of Time Data
• Continuous – a series of
values that change over time
Non-periodic
Periodic
Time Value
10:00 57
10:01 62
10:02 60
10:03 60
etc. etc.
• Discrete – an event that
occurs at a specific time
Non-periodic (non-recurring)
Periodic
Time Event
20 July 1995 John born
30 May 2013 John graduates HS
INFOVIS8803DV > SPRING 17
Continuous Time Data
Stock Price
INFOVIS8803DV > SPRING 17
Continuous Time Data
Weather
INFOVIS8803DV > SPRING 17
Continuous Time Data
Heart Rate
INFOVIS8803DV > SPRING 17
Continuous Time Data
INFOVIS8803DV > SPRING 17
INFOVIS8803DV > SPRING 17
Discrete Time Data
• Birth/marriage/death
• Historical events– Natural disasters
– Flight departures
• Some discrete events have duration (start, stop)– Meeting or appointment
– Hospital stay
– Project phase
– Person’s lifespan
• Some discrete events are periodic– Presidential elections (every 4 years)
– Our class meetings (every Tue and Thu)
– Other examples?
INFOVIS8803DV > SPRING 17
Time Series Tasks:
What are some things we might want to
learn from time series?
INFOVIS8803DV > SPRING 17
Time Series Tasks
• Find highs and lows
• When did X happen?
• Did A happen before or after B? Or did they occur simultaneously?
• How long did it take?
• When will it happen again?
• How often did X occur?
• Determine periodicity
• Compare two time series
INFOVIS8803DV > SPRING 17
Time Series Examples
As we step through the examples, keep these questions in mind
• What are similarities?
• What are differences?
• When is each useful?
• Pros/cons of each?
– How much info is visually encoded?
– Scalable to more events & longer time scale / intervals?
– How good for comparing multiple time series?
INFOVIS8803DV > SPRING 17
INFOVIS8803DV > SPRING 17
Example: Calendar Events on Timeline
well-formed-data.net/experiments/folding_time/
INFOVIS8803DV > SPRING 17
Example: The Meeting InfoVis
INFOVIS8803DV > SPRING 17
Individual Agenda Item
INFOVIS8803DV > SPRING 17
The Meeting InfoVis
INFOVIS8803DV > SPRING 17
Example: Personal Histories
• Lives are typically chronological
• How can you compare, document, investigate events?
INFOVIS8803DV > SPRING 17
Lifelines
Plaisant et al CHI ‘96
INFOVIS8803DV > SPRING 17
Electronic Medical Records – non-graphical
INFOVIS8803DV > SPRING 17
Electronic Medical Records – Graphical with Lifelines
INFOVIS8803DV > SPRING 17
Lifelines Features
• Different colors for various event types
• Line thickness corresponds to another variable
• Interaction: clicking drills down
INFOVIS8803DV > SPRING 17
What are the Benefits?
• Provide information such that:
– Likely reduce missing information
– Make it easy to discover trends or anomalies
• Streamline access to details
• Remain simple and customizable for various applications
• Assumes critical information is in the EHR (Electronic Health
Record)!!!
INFOVIS8803DV > SPRING 17
Challenges for Lifelines
• Scalability to thousands of tests etc.
• Comparisons of test data and images across time
– Do the yearly MRI scans show improvement?
– What trends do we see in the last 12 EKGs?
• Comparisons across cases
– How compare the histories of those 12 people with the mysterious
illness?
– Maybe better to compare using computer rather than our eyes/brain?
Or maybe not …..
INFOVIS8803DV > SPRING 17
Example: Find Patterns in TimeLines
• Want to find patients who received contrast agent for medical
imaging and within two weeks had high levels of creatinine
(indicating renal problems).
• More info: http://www.cs.umd.edu/hcil/lifelines2/
INFOVIS8803DV > SPRING 17
INFOVIS8803DV > SPRING 17
Step 1: Unaligned Medical Records
INFOVIS8803DV > SPRING 17
Step 2: Align Event Times
Yellow
triangles
(when
contrast
agent
administered)
all aligned
INFOVIS8803DV > SPRING 17
Step 3: Rank by Creatinine Level
Sort by
Creatinine
Level
Highest at
top (red
triangles)
INFOVIS8803DV > SPRING 17
Step 4: Summarize Across Cases
Histogram
across cases
INFOVIS8803DV > SPRING 17
Step 5: Expand Time Line
Expand
time line to
see more
detail
Video: http://www.cs.umd.edu/hcil/lifelines2/videos/Lifelines2-%28Contrast-
Creatinine%29-11-10-08_flash/full%281024x768%29/index.htm
INFOVIS8803DV > SPRING 17
EventFlow
http://www.cs.umd.edu/hcil/eventflow/
http://hcil.umd.edu/eventflow/#video
http://medianetwork.oracle.com/video/player/2079912021001
INFOVIS8803DV > SPRING 17
Example: Finding Daily Patterns
• Suppose you have a daily log, for a year, of energy consumption in a
building, or number of people in building
– Want to find common characteristics
– Idea applies beyond this example
Wijk and Selow, Cluster and Calendar based Visualization of Time Series Data, InfoVis ‘99
INFOVIS8803DV > SPRING 17
Raw Data
INFOVIS8803DV > SPRING 17
INFOVIS8803DV > SPRING 17
Cluster Display – People at Work
INFOVIS8803DV > SPRING 17
Cluster Display – People at Work
Weekends
Summer Fridays
Take Day off after Holiday
Leave Early Dec. 5
INFOVIS8803DV > SPRING 17
What Cluster Display Shows Us
• Traditional office hours followed
• Most are present late mornings
• Fewer are present on summer Fridays
• Very few people work holidays
• School vacations
• Day after holidays
• Many people leave at 4PM on 12/5
– Very special in The Netherlands – St. Nicholas’ Eve
INFOVIS8803DV > SPRING 17
Connected ScatterPlot
http://www.nytimes.com/interactive/2013/10/09/us/yellen-fed-chart.html?_r=0
INFOVIS8803DV > SPRING 17
Calendar-Based Heatmap
https://bl.ocks.org/mbostock/4063318
INFOVIS8803DV > SPRING 17
Calendar-Based Heatmap
https://www.trulia.com/vis/tru247/
INFOVIS8803DV > SPRING 17
Example: Spiral Display - Periodic Data
• Useful if data follows a
repetitive pattern
• Can reveal periodicity
One loop is one period
• Time line becomes the spiral
Avoid problem of long time
line
INFOVIS8803DV > SPRING 17
Spiral Display
• Example
12 months (spokes)
Value (consumption) coded
by size
Dots scaled so as not to
overlap
Extra spiral on outside
provides some closure
All spiral display figures from Konstan and
Carlis, Interactive Visualization of Serial
Periodic Data, UIST ’98; data set is
chimpanzee food consumption.
INFOVIS8803DV > SPRING 17
Add Third Dimension for More Data
Mini bar-chart at each point
Black line – start of rainy
season
112 food types
Useful?
Two linked
spirals:
2 chimpanzees
group avg size &
max size
INFOVIS8803DV > SPRING 17
Example: Text Themes over Time
• Consider a set of speeches or documents over time
– How represent the flow of ideas and concepts over time?*
• How would you do this?
*This was an InfoVis student contest
INFOVIS8803DV > SPRING 17
Theme RiverHavre et. al. InfoVis ’00
Each ‘theme’ from news stories is a ‘river’
INFOVIS8803DV > SPRING 17
Theme River
• River height (thickness) encodes relative frequency of themes
• Key events overlaid
INFOVIS8803DV > SPRING 17
Theme River = Stacked Graph?
• Yes, with a differential vertical offset
– Makes a world of difference!
INFOVIS8803DV > SPRING 17
Example: Any Variations over Time
• The ThemeRiver motif can be used to depict many different
variations over time.
Box Office Receiptswww.nytimes.com/interactive/2008/02/23/mov
ies/20080223_REVENUE_GRAPHIC.html
Baby Name Wizardwww.babynamewizard.com
INFOVIS8803DV > SPRING 17
Recall Horizon Graphs from Zoo Reading:
Cubism.js Example
https://square.github.io/cubism/demo/
INFOVIS8803DV > SPRING 17
Example: History of Text Revisions
• IBM’s “history flow” shows changes to documents edited by multiple
authors
http://researchweb.watson.ibm.com/history/
INFOVIS8803DV > SPRING 17
Decoding History Flow
Connect author’s
contributions
from revision to
revision
Time or revision number
Po
sitio
n o
f te
xt
in
do
cu
ment
Length of each
segment =>
amount of text
Show each
author’s
contribution to
each revision
INFOVIS8803DV > SPRING 17
History Flow Pictures
Spacing by time
Spacing by revision #
• Brightness indicates text age
• Registered authors color-
coded
• Anonymous authors in white
INFOVIS8803DV > SPRING 17
Example: Querying
• Most systems focus on visualizing and navigating time-series data.
• What about querying?
INFOVIS8803DV > SPRING 17
TimeFinder
Hochheiser & Shneiderman
Proc. Discovery Science ’01
Info Vis ‘04
Can create rectangles
that function as matching
regions
Light gray is all data’s extent
Darker grayed region is data
envelope that shows extreme
values of queries matching
criteria
INFOVIS8803DV > SPRING 17
TimeFinder – Other Capabilities
• Search for matches based on angle (slope) ± tolerance
• “Leaders and laggards”
– Find same patterns but shifted in time
Slope
Tolerance
INFOVIS8803DV > SPRING 17
TimeFinder Limitations
• Hodgkins patients exhibit double spike in temperature…
– But with differing times between spikes
INFOVIS8803DV > SPRING 17
Example:TimeSearcher
• Overcomes TimeFinder limitation
• Allow time boxes with deltas on each side (TimeSearcher) Video:
2005_timesearcher.mpg
Buono, Plaisant, Simeone, Aris, Shneiderman, et.al, Similarity-Based Forecasting with Simultaneous Previews: A
River Plot Interface for Time Series Forecasting, 11th International Conference on Information Visualization, 2007
INFOVIS8803DV > SPRING 17
TimeSearcher
INFOVIS8803DV > SPRING 17
Example: Query by (Graphical) Example
• Sketch the Query
• Will be discussed in Interaction Lecture
– Specify a timeline query by drawing a rough pattern for it and the
system returns similar matches. (Wattenberg, CHI ‘01)
User-drawn
query
INFOVIS8803DV > SPRING 17
Example: Project Management
• Project (write software, design/build plane, hire new person)
involves
– Multiple steps
– Spread over time
– Some steps depend on other steps
• How can we plan/manage project?
• With PERT/GANTT/CPM charts, of course.
INFOVIS8803DV > SPRING 17
PERT Chart - Time Dependencies
• PERT = Project Evaluation and Review Technique
INFOVIS8803DV > SPRING 17
PERT leads to GANTT
• Dependencies on timeline
INFOVIS8803DV > SPRING 17
Revised GANTT
Shows
- Resources (colors)
- Critical path activities
INFOVIS8803DV > SPRING 17
Revised GANTT Chart Tells Us That
• Ten week completion IF all goes well
• Completion in ten weeks requires:
– 1 analyst for the first 5 weeks
– 1 programmer for 5 weeks starting week 4
– 1 programmer/QA expert for 3 weeks starting week 6
• Analysis, development and
testing of supporting
modules must be on time.
• Hardware not time-critical
but must complete before
Core Module Training start
From
http://www.mindtools.com/pages/article/newPPM_03.htm
INFOVIS8803DV > SPRING 17
Many Project Management Products!
Microsoft Project
INFOVIS8803DV > SPRING 17
Storylines
Sister & Alice together from t = 0
for 4 time units
Design
Considerations for
Optimizing Storyline
Visualizations,
Yuzuru Tanahashi
and Kwan-Liu Ma,
IEEE Transactions
on Visualization and
Computer Graphics,
Dec. 2012
INFOVIS8803DV > SPRING 17
• Lines representing interacting characters must be adjacent.
• Otherwise, lines must not be adjacent.
• A line must not bend except to converge or diverge with another line.
https://xkcd.com/657/
INFOVIS8803DV > SPRING 17
Timeline Curator
http://www.cs.ubc.ca/group/infovis/software/TimeLineCurator/http://www.cs.ubc.ca/group/infovis/software/TimeLineCurator/#video
INFOVIS8803DV > SPRING 17
Timeviz.net Browser
http://timeviz.net/
INFOVIS8803DV > SPRING 17
HW4: Timeline
• Convert your professional CV/Resume into a Visualization
– Consider Events of Interest (Education, Certification/Courses, Employment, etc.). Make sure to include at least 4 categories and multiple events for each.
– Include both Discrete and Continuous Events
– Please be as detailed as you can be (you may exclude events that are personal and that you don’t want to share with others)
– Create a visual timeline dashboard of your CV.
Option 1: You may sketch your timeline (which may allow you to be really creative)
Option 2: Implement the visual timeline in Tableau (I realize there are limitations, but be as creative as you can be).
• Bring two (2) copies of timeline to class and submit HW4 on T-Square.