Re: clustering algorith

[Darren Schreiber]

My coalition formation model is really just a souped-up clustering 
algorithm.  I have been meaning to update my webpage with the cleaned, 
Swarm-2.1 compliant, and repeating (a la Paul Johnson) code.  If I can, I 
will try to get that up this weekend, but I'm making final preparations for 
the American Political Science Association conference, so no promises.

There are a wide variety of clustering algorithms out there, but someone 
has shown that most are mathematical transformations of each other.  The 
algorithm I use is centroid-linkage, because it was easy to program and has 
a substantive fit with the story I am telling.  In my model, each voter 
looks to find its closest neighbor.  If two voters agree that they are each 
others closest neighbor, then they form a coalition (aka cluster) that is 
located at the centroid of its constituent voters.  The two voters are 
dropped out of the list of available agents and the coalition is put into 
the list.  The cycle repeats until a hierarchy of coalitions is built.  The 
top level coalition contains all other coalitions and all the voters.

I have also used clustering algorithms for data analysis with some of my 
political cognition research and found it can be helpful.  Sometimes, I got 
great results, other times...  If you are only wanting to do data analysis, 
I think many of the canned programs can do it for you.  I know SPSS and 
Data Desk both do a good job with it.

I can give you a pretty good bibliography when I get home.

        Darren



At 03:11 PM 8/25/00 -0500, you wrote:
> Can you point me at C/objc-c code that I can use to decide if there are
> separate clusters in a set of points in 2D space and then to sort the
> points among the clusters if they exist?
>
> Maybe you have to tell me if clustering is workable, or where to learn
> more about it.
>
> --
> Paul E. Johnson                       email: address@hidden
> Dept. of Political Science            http://lark.cc.ukans.edu/~pauljohn
> University of Kansas                  Office: (785) 864-9086
> Lawrence, Kansas 66045                FAX: (785) 864-5700
>
>
>                   ==================================
>    Swarm-Modelling is for discussion of Simulation and Modelling techniques
>    esp. using Swarm.  For list administration needs (esp. [un]subscribing),
>    please send a message to <address@hidden> with "help" in the
>    body of the message.
>                   ==================================

_______________________________________

           Darren Schreiber
           Attorney at Law
          Graduate Student
      Political Science, UCLA
          address@hidden
  http://www.bol.ucla.edu/~dschreib



                 ==================================
  Swarm-Modelling is for discussion of Simulation and Modelling techniques
  esp. using Swarm.  For list administration needs (esp. [un]subscribing),
  please send a message to <address@hidden> with "help" in the
  body of the message.
                 ==================================