Skip down to main content

New Pinwheel network layout

Published on
10 Jan 2010
Written by
Bernie Hogan

I’m presently working on a chapter on Facebook networks for the forthcoming Analyzing Social Media Networks with NodeXL edited by Marc Smith, Derek Hansen and Ben Shneiderman. Working with [NodeXL][nx] has been a lot of fun. Since NodeXL does not have native facilities for importing Facebook, there’s not much to discuss for data collection. I simply describe my tool for exporting edge lists and paste them into the Excel worksheet. Consequently, I spend a lot of time showing how to push out new and interesting layouts.

One of the layouts I show people how to create is something that looks like [Thomas Fletcher’s friend wheel][tf]. I’ve always enjoyed the look of a Friendwheel but I don’t find it especially informative. To beef up this visualization, I toyed around with NodeXL to get a new visualization that I call a Pinwheel.

Friendwheel-esque layout by NodeXL

Pinwheel via NodeXL

After the cut I describe some details of the visualization and how to interpret it.


Each wing of the Pinwheel represents a group as discovered by a community detection algorithm. The head of the pinwheel is the node of highest degree and the tail includes nodes of smallest degree. The colors and sizes are also mapped to degree to reinforce the overall shape of each wing. The nodes closest to the center are nodes of highest betweenness. I chose betweenness to reflect a sort of ‘gravity’ where these nodes want to be pulled towards other groups. I use an opacity of 50% for both the nodes and the edges so you can see overlapping nodes and edges.

I like this diagram for several reasons:

* The angle of each wing (i.e. each cluster) is proportionate to its share of the network. So if a cluster has 25% of all nodes, it will have 90 degrees.

* Because of the shading and the arcs (through betweenness) each wing does not need to be colored differently in order to be seen as a distinct group (which is nice when it has to be printed in grayscale).

* By looking at the tail you can see how much of a periphery each group has. Those groups with many small nodes at the tail have lots of members who only know people in that group.

* By looking at the head you can see which groups have people with the highest overall degree.

* By looking at the edges that cross between the wings you can see which groups are well connected, but also because the nodes are sorted by degree it is easy to see if the best connected nodes are the ones that mainly link the groups, or if the entire group links from one wing to another.

* Finally, if each wing goes from white to blue it looks like an iceball and if each wing goes from red to yellow it looks like a fireball. (Can you tell I’ve been playing a little too much Super Mario Bros. Wii?)

Related Topics