[R] Compare clustering solutions to a "correct" one

Andrew Ziem AZiem at us.ci.org
Wed Nov 9 21:07:14 CET 2011

It sounds like you want to do supervised classification, so maybe a supervised classification algorithm would be more appropriate?  Consider logistic regression, rpart, ctree, earth, etc.


-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Massimo Lole
Sent: Wednesday, November 09, 2011 12:39 PM
To: r-help at r-project.org
Subject: [R] Compare clustering solutions to a "correct" one

Hello everyone,
I have a set of data, J healthy subjects, K diseased subjects, N features for each person. I would like to clusterize my data. Since I know that subjects are from two populations ideally I would prefer an algorithm that first is able to discriminate them, in order to see how it performs inside each group.
I know that R offers several clustering functions and a very interesting clustering compare tool: cluster.stats in fpc package.
I would like to ask you if you know of any existing approach where one cluster is considered "correct" and against it several clustering functions (with several parameters) are run, benchmarking them and selecting the best one.

Since I am new of R a skeleton procedure with useful functions would help me a lot in setting up this test.

Thank you very much,

More information about the R-help mailing list