Friday, November 21, 2008

Geographic networks

Geography (or spatial arrangement) of nodes is often an important factor in network dynamics. Though it is straightforward to map geographical information by itself, mapping that information simultaneously with network data is quite a challenge.

In collaboration with Holly Massett and her team at the National Cancer Institute, I have been tackling the geographic + network mapping problem head on. Holly and I recently presented some of our results, and she graciously gave me permission to share them.

What happens when we draw a network map with geographically located nodes? We get a map with lines on it:
The geography is plainly apparent, but the network structure is all but invisible. That's a shame, because the network structure hidden above is actually quite striking when you redraw the above network using traditional network layout techniques:
Now we can clearly see that there is one node that bridges between two distinct clusters.

As a simple first step toward integrating these two important views of the above collaboration network, I created this slide show, which morphs back and forth between pure geography and pure network information, showing the interaction of the two along the way (RSS readers must view my actual blog to see this):

This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 License and is copyrighted (c) 2008 by Connective Associates LLC except where otherwise noted.


Aftersox said...

I'm sure there's a good way of visualizing geographic network data. Maybe a cartogram of sorts? A bipartite network? Weight the distances between nodes based on geographic distance somehow?

The problem is that social networks are very weakly connected to geography. I don't know of any papers that try to resolve this... low hanging fruit anyone?

Alan Glennon said...

To visualize the network interconnections within a geographic framework, one possibility is to use graduated node sizes (varying by the number of input edges), line coloration or differing widths (varying by hierarchical order).

seth said...

In response to Aftersox's comment about the connection between networks and geography, I recently came across a very interesting paper that does try to address this, How relevant are features for network structure?" by Ginestra Bianconi, Paolo Pin, Matteo Marsili.

One of the applications of the new measure they create is to measure the connection between the network of airplane flights and the geographical position of airports. I wonder if anyone has seen this paper or has thoughts on it?