Mining the Social Web teaches you how easy it is do to things exactly like this and provides template code to get you started on your way. It's much simpler than you might think, and can actually be fun.

All of the data that you need is available to you (for your own networks) through APIs, and munging it into a suitable data structure for a map visualization. From a data-mining and programming standpoint, it's a fairly trivial exercise, but can be pretty fun to see the results for yourself.

See chapters 2+3 of Mining the Social Web (2nd Edition) for some helpful starting points if this is the kind of project you'd like to try out for a new project - http://amzn.to/GPd59m

Incidentally, the 1st Edition of Mining the Social Web (from Feb 2011) provided a faceted display that was essentially this exact thing - a bipartite graph of where friends live now versus their hometown. Although I do not recommend that anyone use the 1st Edition (now defunct), the sample code for this exercise was here - http://bit.ly/19Ni4mE

It's easy enough to adapt, and I'd be glad to help anyone who is interested.

Thank you Matthew!

This is a good use of Facebook data.

As opposed to Facebook using private messages to target ads:

http://businessinsider.com/facebook-sued-for-allegedly-using-your-private-messages-to-trigger-ads-2014-1

Facebook should stick to good uses of its data unless they want a reputation for doing bad things.