1

Information VisualizationInformation Visualization

Ji Y

1

Jing YangSpring 2010

Time Series Data Visualization

2

2

Time Series Data

Fundamental chronological component to the data setdata setRandom sample of 4000 graphics from 15 of world’s newspapers and magazines from ’74-’80 found that 75% of graphics published were time series

Tufte

3

− Tufte

From John Stasko’s class slides

Datasets

Each data case is likely an event of some kindkindOne of the variables can be the date and time of the eventExamples: sunspot activity, baseball games, medicines taken, cities visited, stock prices, newswires network resource measures

4

newswires, network resource measures

Partially From John Stasko’s class slides

3

DiscussionVisualize the following datasets: (1) CO2 emission(1) CO2 emissionDimension 1: Year Dimension 2: Country Dimension 3: Income per personDimension 4: CO2 emissions per personDimension 5: Population

5

(2) News Each news article has a time stamp, a list of

keywords describing their contents, a title, and a link to its original content

Time Series Visualization Approaches

Static State Replacement (Animation)S ll M lti lSmall MultiplesTime-Series PlotNested Visualization (embed time-series plot into other display)Brushing and linking

6

4

Small MultiplesSmall multiples are sets of thumbnail sized graphics on a single page that represent aspects of a single h Thphenomenon. They:

Depict comparison, enhance dimensionality, motion, and are good for multivariate displays Invite comparison, contrasts, and show the scope of alternatives or range of options Must use the same measures and scale. Can represent motion through ghosting of multiple images

7

images Are particularly useful in computers because they often permit the actual overlay of images, and rapid cycling.

Graphics and Web Design Based on Edward Tufte's Principles, Larry Gales, Univ. of Washington

Small Multiples

8Three air pollutants in six counties in southern California Los Angeles Times, 1979

5

Shape Coding

9Beddow J.: ‘Shape Coding of Multidimensional Data on a Mircocomputer Display’, Visualization ‘90, 1990, pp. 238-246.

Static State Replacement

Treat time as a dimension hidden from the displaydisplayDivide time into period (time frame, or time point)Generate a visualization for each time frameReplace a display of one time frame using th t f th ti f

10

that of another time frameAnimations, trails

6

Gapminder Trendalyzer

Animated bubble chart to show trends over time in three dimensionstime in three dimensions

Hans Rosling: No more boring data. TED (Technology,

Entertainment, Design) 2006Video

Gapminder WorldDemo

11

Change BlindnessPeople do not notice changes in visible elements of a scenePossible reasons:

OverwritingOld scene is wholly replaced by the new one

First impressionsAccurately encode details of first scene and fail to encode the details of the changed scene

Nothing is storedNo need to develop any mental representation of the scene

Nothing is compared

12

Nothing is comparedNeed to focus on changed items to recognition of changes

Feature combinationNew scene and old scene are combined together

Nowell et al. Infovis 01

7

Example

13

VideoGalaxies slices depicting days 1-3

Nowell et al. Infovis 01

Change Blindness

14

VideoThemeview slices depicting days 1-3

Nowell et al. Infovis 01

8

Paper: Change Blindness in Information Visualization: A Case Study [Nowell et al. Infovis01]

Portraying document age i G l i Vi li tiin Galaxies VisualizationRequirements:

Relative age should be apparentNewest documents to be

tt ti l

15

seen pre-attentivelyOther document ages to be intuitively ordered

Paper: Change Blindness in Information Visualization: A Case Study [Nowell et al. Infovis01]

Check pre-attentive features:Spatial layoutSpatial layout Size Shapes Angles Line lengthColor progression (such as yellow to green to blue)

16

Bright to dim progressionPerspective depthLeft to right spatial progression

9

Perspective depth, line and length encoding

17

Line angle and length solution

18

10

Paper: Change Blindness in Information Visualization: A Case Study [Nowell et al. Infovis01]

Candidate solutions for ThemeViewThemeView

MorphingWhat come before, what will eventually appear?Does not help users remember the changes

Cross-fadingWhich part will get brighter

19

Which part will get brighter, which part will fade away?

Using a wireframe in combination with changes in color and translucency

Wireframe Solution

Moving from one time slice to another with a wireframe and variable translucency.

20

g y

11

Theme Scan Solution

21

ThemeScan visualization of changes between time slices

Trace

Epic Systems Trend Compass animated charts sample is on Monitoring ads on TVcharts. sample is on Monitoring ads on TV channels

http://www.epicsyst.com/ads3/allchannels.swf

22

12

User study

Effectiveness of animation in Trend visualizations G Robertson et al Infovis 08visualizations G. Robertson et al. Infovis 08

23

Time Series Plot

24

Inclinations of the planetary orbits as a function of time

Part of a textbook of monastery schools, tenth century

13

Time Series Plot

25

Time Series Plot

26

14

Time Series Plot

27

Time Series Plot

28

15

TimeSearcher

Free software from HCILF tFeatures:

Visualize multivariate long time seriesProvide overview that can be zoomed inSupport interactive pattern searchSupport search by example patterns

VideoDataset: eBay auction data (prices, velocity) etc.

29

Spiral Graphs

30History of Italian post office A. Gabaglio, 1888

16

Paper: Visualizing Time-Series on Spirals [weber et al. Infovis 01]

31

Features

Scale to large data setsS t id tifi ti f i di t t iSupport identification of periodic structures in the dataCompare multiple datasetsUse Archimedes’ spiral: r = aӨ

A ray emanating from the origin crosses two

32

consecutive arcs of the spiral in a constant distance 2πa (equal distance between adjacent periods)

17

Periodic Pattern Identification

Spectrum analysisA i tiAnimation

33

Multiple Spirals

34

18

Scales & Legends

35

3D Overview and Selection

36

19

Pixel-Oriented Techniques

Recursive pattern arrangements

37

The figure is taken from Dr. D. Keim’s tutorial notes in Infovis 00

Pixel Oriented Techniques

Recursive pattern arrangements

38

The figure is taken from Dr. D. Keim’s tutorial notes in Infovis 00

20

Nested Visualization

Embed time series plot into other displaysE l Ti i l t b dd d i t hExample: Time series plot embedded into a graph

39

Visualization of Graphs with Associated Timeseries Data [Saraiya:05]

Brushing and Linking

Link time series display with other displays

40

Visualization of Graphs with Associated Timeseries Data [Saraiya:05]

21

Space and Time

41Napoleon’s army in Russia, author: Charles Minard (1781-1870)

Space and Time

42

Life circle of Japanese Beetles L. Newman, Man and Insects, 1965

22

Paper: GeoTime Information Visualization [Kapler and Wright Infovis 04]

A combined temporal-spatial space (X, Y, T coordinate space)coordinate space)Represent place by 2D plane (or maybe 3D topography)Use 3rd dimension to encode time

43

Paper: GeoTime Information Visualization [Kapler and Wright Infovis 04]

44

23

Timelines

3-D Z axis timelines 3-D viewer facing timelinestimelines

45

Example

46

24

Example

47

Information Model

Entities P l thiPeople or things

Locations Geospatial or conceptual

Events

48

Occurrences or discovered facts

25

Association Analysis

Expanding search Connection search

49

Other Interactions

Animation of entity movements

AnnotationsmovementsDrilling down

50

26

Alternative View

Afghanistan in 2002E t i th kEvents in three weeks

ShootingsBombingsFiresMines

51

KidnapsTheftsassaults

GeoTime: H1N1 "Swine Flu" Visual Analysis

htt // t b / t h? Ok hjbhttp://www.youtube.com/watch?v=Okghjb_oABM&feature=related

52

27

Text Collections and Time

Constantly evolving text streamsE h d t h ti tEach document has a time stampTemporal dynamics Alignment problem: find real life events happen in parallel and drive the text

53

Paper: ThemeRiver: Visualizing Theme Changes Over Time [Havre et al. Infovis 00]

VideoB k d i l i t t d iBackground: a user is less interested in document themselves than in theme changes within the whole collection over timeThemeRiver provides users a macro-view of thematic changes E l d t t d

54

Example dataset used: 1990 Associated Press (AP) newswire data

28

A histogram depicting thematic changes

55

Problem

The position of a particular theme within the bars may vary considerablybars may vary considerablyUsers are required to integrate the themes across time Improvement: the river and currents metaphor -> ThemeRiver

56

29

ThemeRiver

The river flows from left to right through timeC l d t fl i ith th iColored currents flowing with the river narrow or widen to depict the strength of individual topics

57

58

30

59

60

31

EventRiver

By my student Dongning Luo

Motivation: help users detect, track, and investigate real life events that drive the text collections

Approach: Visual analysisIncremental text clustering

61

Event Overview

62

32

Search by Keyword

63

Shoebox

64

33

Paper: Time-Varying Data Visualization UsingInformation Flocking Boids [Moere Infovis04]

Motivation: users are interested in how data values evolve in time or in the context of thevalues evolve in time, or in the context of the whole dataset, rather than exact data valuesExample: stock price

A company performing significantly better than the day beforeA company performing significantly better than

65

A company performing significantly better than the day before

Flocking Boids

Boids (bird-objects) within a flockB id t th d f h b i t bBoids at the edge of a herb are easier to be selectedBoids attempt to move as close to the center of the herd as possibleBoids view the world from their own perspective rather than from a global one

66

perspective rather than from a global one

34

Behavior AnimationEach individual member contains its own set of rules and the future state of a member only depends on its y pneighborsRules:

Collision AvoidanceVelocity Matching (move with about the same speed as neighbors)Data similarity (Stay close to boids experienced similar data value evolution during current timeframe)

67

data value evolution during current timeframe)Data Dissimilarity (Stay away from boids experienced dissimilar data value evolution)Flock Centering (move toward the center as the boid perceives it)

Example

68

35

Shape

69

References

E. Tufte. The Visual Display of Quantitative Information 1983Information, 1983Papers referred

70