Network analysis

[beingkk]

Network analysis

Basic progress so far

See 32_network_analysis, [notebook for getting started]:(https://github.com/nestauk/innovation_sweet_spots/blob/32_network_analysis/innovation_sweet_spots/analysis/prototyping/network_analysis/01_network_analysis_intro.py)

• Builds a collaboration network between research organisations working on a technology
  • Nodes = organisations, links = a common project
• Visualises the network using altair (using Juan's code from the Createch project). There are some issues with the visualisation, i.e. the mouse-over doesn't work in the notebook - whereas in the html file, the zooming doesn't work.

Some potential next steps

Using Gateway to Research data

1. Interactivity
   • Better visualisation using Altair, allowing to explore the network, check organisations and what projects form the basis of the links
   • Exploring displaying these networks in kumu, which is used by the ASF team
2. Performing simple analyses and highlighting:
   • Which organisations are most the central ones? (i.e., most interconnected, hinting at their importance)
   • How different types of organisations (universities, businesses, local authorities) are placed in the network, how central they are? Here the challenge is, however, that organisations don't have designated types and we would have to somehow infer this from their names.
   • Perhaps: Communities or connected components of organisations. Maybe a nearest-neighbour subgraph (involving all direct neighbours) is interesting to highlight
   • Also: Can we explain the results (e.g. why a particular organisation might appear very central?)
3. More advanced analysis: Characterising network growth over years
   • Visualising the new links across years
   • Can we say something about a 'growth mechanism' ? (e.g. does it look like preferential attachment, i.e. rich get richer, or more like random growth)

It would also be very interesting to use Crunchbase data to visualise the network of investors and businesses

We can also keep in mind geographical analysis (as organisations also have address information). This would, however, be a task for another issue.