hierarchical techniques

Transcription

1 Wolfgang Aigner Version hierarchical techniques WOLFGANG AiGNER hierarchical techniques 1

2 Overview A) Hierarchical visualization techniques B) Visualizing hierarchical data [Keim, 2001] WOLFGANG AiGNER hierarchical techniques 2

3 Part A hierarchical visualization techniques WOLFGANG AiGNER hierarchical techniques 3

4 Problem Data multivariate data (e.g., Movie DB, Car dataset) Presentation space display dimensionality constrained to 2D or 3D Task meaningful representation of all variables within a single plot How? WOLFGANG AiGNER hierarchical techniques 4

5 Example 4 variables: longitude latitude ore grade depth WOLFGANG AiGNER hierarchical techniques 5

6 Dimensional Stacking [LeBlance et al. 1990] WOLFGANG AiGNER hierarchical techniques 6

7 Dimensional Stacking [LeBlance et al. 1990] Partitioning of the n-dim. attribute space in 2-dim. subspace, which are stacked into each other Partitioning of the attribute value ranges into classes Important attributes should be used on the outer levels Adequate especially for data with ordinal attributes of low cardinality WOLFGANG AiGNER hierarchical techniques 7

8 Dimensional Stacking [LeBlance et al. 1990] Given Variables V 1 - V m Power of the Domain: Cardinalities K 1 K m Process 2 Variables V i,v k K i *K k Grid V k V i K i K k WOLFGANG AiGNER hierarchical techniques 8

9 Dimensional Stacking [LeBlance et al. 1990] Recursive Repetition with a Grid with Additional Pairs of Variables V k V m... V i K i V n K n K k K m Dummy-Variable: 2n-1 Variables WOLFGANG AiGNER hierarchical techniques 9

10 Example [LeBlance et al. 1990] Variable: V 1 -V 6 Numbers of Cardinality: K 1 =4,K 2 =2,K 3 =2,K 4 =3,K 5 =3,K 6 =2 WOLFGANG AiGNER hierarchical techniques 10

11 Example [LeBlance et al. 1990] Variables: V 1 -V 6 Cardinalities : K 1 =4,K 2 =2,K 3 =2,K 4 =3,K 5 =3,K 6 =2 Pairs P 1 (V 1,V 3 ),P 2 (V 4,V 5 ),P 3 (V 2,V 6 ) Example - Combination: 4,2,3,2,2,2 WOLFGANG AiGNER hierarchical techniques 11

12 Worlds-within-Worlds [Feiner & Besherss 1990] Partitioning of the n-dim. Space into 3-dim. Subspace Nested 3-dim coordinates 3-dim coordinate = one property Selected points --> new coordinates system WOLFGANG AiGNER hierarchical techniques 12

13 Worlds-within-Worlds [Feiner & Besherss 1990] WOLFGANG AiGNER hierarchical techniques 13

14 Part B visualizing hierarchical data WOLFGANG AiGNER hierarchical techniques 14

15 Basic Data Characteristics network / graph items (nodes) that have relationships (links) no inherent hierarchical structure hierarchical / tree parent - child relationships every node has at most one parent exactly one root node mostly non-leaf nodes are containers only WOLFGANG AiGNER hierarchical techniques 15

16 Data Hierarchical data are very common Hierarchies are one of the most prevalent organizing principles for coping with information application examples organizations, org-charts, taxonomies, table of contents, sitemaps, file system, genealogies,... WOLFGANG AiGNER hierarchical techniques 16

17 Tasks reducing complexity categorization - hierarchies (expand/collapse) overview of topology distribution examine relationships examine paths examine elements identify locate distinguish relate compare specific general WOLFGANG AiGNER hierarchical techniques 17

18 Illustrating example Data: file system Problem/task: disk is full --> free some space How? Tool? Visualization? Tjark Derlien, Disk Inventory X, WOLFGANG AiGNER hierarchical techniques 18

19 Visual Encodings indentation representation of hierarchy level via indentation focus on linear structure connection node-link diagrams convention: root mostly on top, leafs on bottom aspect ratio containment summed values propagation through hierarchy space-filling graphs adjacency matrices graph as table nodes as rows/columns edges as table cells Dept-A Dept-B Center-1 Dept-C Dept-E Dept-D Center-2 Company-X Dept-F Center-3 Images 1 and 2: Steven F. Roth and Joe Mattis, Data Characterization for Intelligent Graphics Presentation, CHI'90, WOLFGANG AiGNER hierarchical techniques 19

20 Node-link diagrams Graph drawing huge area lots of theory and algorithms Good layout is important for effective presentation of inherent structure basic layouts linear tree radial force directed spring gravity magnetic Links - Gestalt law - line smoothness Images: - Bernard J. Kerr, Thread Arcs, 2003, s.html. - Jeffrey Heer, Tree Visualization, SIMS 247: Information Visualization and Presentation, Max Baker, Netdisco, WOLFGANG AiGNER hierarchical techniques 20

21 Interaction why? aspect ratio large information space do not fit onto display space Problem: large structures that don't fit on a single view/screen expand / collapse navigate focus + context see upcoming lecture for details WOLFGANG AiGNER hierarchical techniques 21

22 visualization techniques WOLFGANG AiGNER hierarchical techniques 22

23 Indented Lists representation of hierarchy level via indentation focus on linear structure WOLFGANG AiGNER hierarchical techniques 23

24 SpaceTree / DOI Tree Demo [Plaisant et al., 2002] [Heer and Card, 2004] WOLFGANG AiGNER hierarchical techniques 24

25 Cone Trees [Robertson, Mackinlay, Card 1991] WOLFGANG AiGNER hierarchical techniques 25

26 Cone Trees vs. Cam Trees Vertical (Cone Tree) vs. Horizontal (Cam Tree) Video Shadows provide 2D structure WOLFGANG AiGNER hierarchical techniques 26

27 Cone Trees [Robertson, Mackinlay, Card 1991] Important: Interaction! WOLFGANG AiGNER hierarchical techniques 27

28 Starlight File System [Pacific Northwest National Laboratory USA] WOLFGANG AiGNER hierarchical techniques 28

29 Balloon Trees Flattened cone trees [Herman, Melancon, and Marshall, 2000] WOLFGANG AiGNER hierarchical techniques 29

30 Hyperbolic Trees Nodes are placed on hyperbolic geometry (inside of a sphere) [Munzner, 1998] Demo Projection into 2D F+C WOLFGANG AiGNER hierarchical techniques 30

31 FAS.research Social Network Analysis WOLFGANG AiGNER hierarchical techniques 31

32 Containment [Shneiderman 1992; Johnson, 1993] Venn- Diagram WOLFGANG AiGNER hierarchical techniques 32

33 InfoCube [Rekimoto & Green 1993] 3 D Visualization of Hierarchical Data Using Transparent Boxes WOLFGANG AiGNER hierarchical techniques 33

34 Application example: Clustering search engine Vivisimo Indented list WOLFGANG AiGNER hierarchical techniques 34

35 Application example: Clustering search engine Grokker Containment hierarchy WOLFGANG AiGNER hierarchical techniques 35

36 Venn-Diagram --> Treemaps [Shneiderman 1992; Johnson, 1993] Nested Treemap Treemap: WOLFGANG AiGNER hierarchical techniques 36

37 Example: File Structure to Tree File System: 3 Folders 6 Files 1) Root -> whole Screen Root Root Dir 1 File 1 File 2 Dir 2 File 3 Dir 3 File 4 File 5 File 6 1 MB 2 MB 2 MB 3 MB 1 MB 1 MB WOLFGANG AiGNER hierarchical techniques 37

38 Example: File Structure to Tree File System: 3 Folders 6 Files 2) Cutting - according to the size (30% and 70% of the space) Root Dir 1 Dir 2 Root Dir 1 File 1 File 2 Dir 2 File 3 Dir 2-1 File 4 File 5 File 6 1 MB 2 MB 2 MB 3 MB 1 MB 1 MB WOLFGANG AiGNER hierarchical techniques 38

39 Example: File Structure to Tree File System: 3 Folders 6 Files 3) Iteration: folder and subfolder File 1 File 1 File 3 Root Dir 1 File 1 File 2 Dir 2 File 3 Dir 2-1 File 4 File 5 File 6 1 MB 2 MB 2 MB 3 MB 1 MB 1 MB File 2 Root Dir 2 Root File 2 Dir 2-1 WOLFGANG AiGNER hierarchical techniques 39

40 Example: File Structure to Tree Demo File System: 3 Folders 6 Files One Solution Root Dir 1 File 1 File 2 Dir 2 File 3 Dir MB 2 MB 2 MB File 1 File 2 Root File 4 File 3 File 5 File 6 File 4 File 5 File 6 3 MB 1 MB 1 MB WOLFGANG AiGNER hierarchical techniques 40

41 Treemap: View Large Trees with Node Values [Shneiderman talk] + Space filling + Space limited + Color coding + Size coding - Requires learning TreeViz (Mac, Johnson, 1992) NBA-Tree (Sun, Turo, 1993) Winsurfer (Teittinen, 1996) Diskmapper (Windows, Micrologic) Treemap3 (Windows, UMd, 2001) (Shneiderman, ACM Trans. on Graphics, 1992) WOLFGANG AiGNER hierarchical techniques 41

42 Finance Analysis Gainers (bright green) WOLFGANG AiGNER hierarchical techniques 42

43 Finance Analysis Losers (bright red) WOLFGANG AiGNER hierarchical techniques 43

44 WOLFGANG AiGNER hierarchical techniques 44

45 WOLFGANG AiGNER hierarchical techniques 45

46 Treemap: Product catalogs WOLFGANG AiGNER hierarchical techniques 46

47 Treemap: Newsmap WOLFGANG AiGNER hierarchical techniques 47

48 TreeMaps Summary Turning a tree into a planar space-filling map Capacity to see tens of thousands of nodes in a fixed space and find large areas or duplicate directories is very powerful Treemap algorithms BinaryTree Ordered SliceAndDice Squarified Strip Beamtree Map of the market [Wattenberg, smartmoney.com] WOLFGANG AiGNER hierarchical techniques 48

49 Icicle Trees Tree levels side by side horizontal / vertical Subdivision by size Demo Randelshofer, WOLFGANG AiGNER hierarchical techniques 49

50 Sunburst Tree [Stasko] Radial version of icicle trees Interaction facilities to navigate / zoom Demo Video Randelshofer, WOLFGANG AiGNER hierarchical techniques 50

51 Botanical Visualization of Huge Hierarchies [Kleiberg, van de Wetering & van Wijk, 2001] Node and link diagram Holton s Strang Modell WOLFGANG AiGNER hierarchical techniques 51

52 Botanical Visualization [Kleiberg, van de Wetering & van Wijk, 2001] Alternative 3D Visualization to Big Hierachies Branches Clash Seldom, Even Though no Particular Algorithm is Included Adapted Phi-Balls are Appropriate fir Big Files WOLFGANG AiGNER hierarchical techniques 52

53 Botanical Visualization: Summary [Kleiberg, van de Wetering & van Wijk, 2001] [Slides from Novotny, 2003] Alternative 3D Visualization to Big Hierachies Branches Clash Seldom, Even Though no Particular Algorithm is Included Adapted Phi-Balls are Appropriate fir Big Files WOLFGANG AiGNER hierarchical techniques 53

54 Summary Hierarchical visualization techniques Re-usage of display dimensions Visualization of hierarchical data Hierarchical / tree data Parent/child relationships Widespread and common data structure Representations Indented lists Node-Link diagrams Containment diagrams Adjacency matrices WOLFGANG AiGNER hierarchical techniques 54

55 Useful Stuff Treemap HCIL Treemap Browser < Map of the Market < Newsmap < The Hive Group < HyperTree Java Library < SpaceTree < Tree Visualizer < VisualComplexity.com < ManyEyes < Search Engines / Clustering Vivisimo < Grokker < WOLFGANG AiGNER hierarchical techniques 55

56 Acknowledgements Thanks to Silvia Miksch whose slides form the basis of this presentation. Ideas have been taken from Katy Börner s, Jeff Heer s, and Jock Mackinlay s presentation slides of their visualization classes. WOLFGANG AiGNER hierarchical techniques 56