Week 4

Reflection

• Time Granularity: should not be limited to natural time division like days, weeks, etc.. Many datasets only use a snapshot id based on equal intervals. So there should be freedom to aggregate the graph based on id or time.

• Multi-layer Timeline: plotting lines together seems not a good idea because of the scale issue. Fine-granularity lines are hard to see.

• Scatter plot view is not suitable for "count" based measures, especially on a small scale. For there will be many overlaps. (0, 0) is often the case and the curve is not obvious.

Good example: Diplomatic Exchange

Bad example: Scientist

• Grouping

  • Selection: instead of selecting nodes on the node-link diagram, it is more intuitive to select on a ranked list or brush on a measure distribution snapshot.

  • Abstraction: Combine nodes in a group to a (collapsible) super node and create a view for group-group interaction. ~ What about on screen touch manipulation?

  • Measure Comparison: parallel coordinate.

• Scalability Issue: for larger graphs, real-time calculation for every local measures is not practical

Exploration

• Diplomatic Exchange 2016 (filtered)

  • almost complete graph and the focus is on link type

  • no particular findings on different measures, without link type information every country appears similar

• Scientist

  • many nodes have poor story

• AVES-WEAVER-SOCIAL: networks represent social data collected from 23 colonies of sociable weavers (src)