Overview of relationships between groups, removing highly redundant groups
A month ago I posted an updated overview of relationships between groups. It was an informative graph, but soon after posting it I realized the methodology highlighted redundancy at the expense of uniqueness. To try to solve that, I started playing with measures of uniqueness.
I now integrated the algorithms and after tweaking quite a bit... I produced a new overview of group pools. As usual there is a main component (big cluster) that you see here, and a whole series of smaller clusters.
In the interest of clarity, this time the graph includes only 2007 nodes (connected by 3378 edges). You can browse this better here. This graph is still based on the same February 2008 data.
...looks like I could be even more aggressive at removing redundancy.