Tuesday, September 13, 2011

A Community-Detection Problem

Take a look at this article on the proposed winners and losers of boundary changes for United Kingdom constituencies.

The following comment appears in the article: New research suggests the Liberal Democrats could lose a quarter of their seats at the next general election before a vote has been cast, under coalition plans to redraw the political map of Britain.

Other projections have suggested the plans would not have such a devastating effect on Nick Clegg's party, but neither would they hand a significant boost to the Conservatives, as many have assumed.

The truth is, nobody really knows.

Hello people: This is a community-detection problem. The data exists*, so one should be able to find the answer pretty easily. Pretending that nobody can possibly have any clue about the answer is woefully naive.

* However, the data might be in a very annoying format, and compiling it to be in a good format might take some non-trivial effort. Maybe somebody has already done that compilation and can have something that is, e.g., Matlab-ready pretty easily?


Antoine said...

How is that a community detection problem? You just have to redo the counting from the last election according to the new map. Guardian has some data and analysis at http://www.guardian.co.uk/news/datablog/2011/sep/13/boundary-changes-election-results

Mason said...

It's a partitioning problem, and one of my special powers is to convert all problems into graph-partitioning problems. :)

Of course, now that you question this idea, I'm curious whether looking at this with community detection will genuinely give insights that something simpler won't already give (which seems to be your actual point rather than whether one can formulate this problem as one of community detection).

Thanks for the link! It is much appreciated.