Re: Finding The Best Cluster Problem

I think a good word to search on google would be "dendrogram." A dendrogram is a graphical representation that lets you explain the best clusters visually; there are most likely some algorithms for generating dendrograms that have code or pseudocode out there for inspiration.

In general, a dendrogram will connect the closest matches, reorder the set so they're consecutive entries in set X, and collapse these neighbors into a single composite entry and try again. Repeat until the size of set X or the distance between clusters achieves your desired threshold.

--
[ e d @ h a l l e y . c c ]

Comment on Re: Finding The Best Cluster Problem

Replies are listed 'Best First'.
Re^2: Finding The Best Cluster Problem by neversaint (Deacon) on May 17, 2007 at 01:34 UTC
Thanks a lot for the reply. Do you know if there exist any implementation of DENDOGRAM in Perl? I tried CPAN, doesn't seem to find any. --- neversaint and everlastingly indebted.......	[reply]
Re^3: Finding The Best Cluster Problem by clwolfe (Scribe) on May 17, 2007 at 03:21 UTC
A dendrogram is a way of visualizing the results of a clustering algorithm. It's not a way to cluster in and of itself. There is a CPAN module that will help you - it's Algorithm::Cluster, which contains several approaches to your problem. You'll need a bit of background in cluster analysis - cluster analysis is a good starting point. Good luck!	[reply]
Re^4: Finding The Best Cluster Problem by spx2 (Deacon) on May 18, 2007 at 11:41 UTC
maybe there's a solution in using linear programming	[reply]