[R] Package cluster

Martin Maechler maechler at stat.math.ethz.ch
Fri Jan 10 16:40:07 CET 2003

>>>>> "Francisco" == Francisco Junior <fnj at cin.ufpe.br>
>>>>>     on Fri, 10 Jan 2003 10:37:02 -0300 (BRST) writes:

    Francisco> Hello,
    Francisco> Someone use this package?

(I hope, it's a recommended package for a reason).

    Francisco> I would like to know if this package has some method to classify
    Francisco> images?

you mean "cluster" or "classify" ?
For many of us "classify" is `supervised' where as
"clustering" is `unsupervised'.

If the second:  
- About how many clusters do you want?
- about how many images do you have?
- what's their format?

In (supervised) classification, this has been quite a hot topic.
In any case, many things depend on your goal and data.
For clustering, you have to think about a (dis)similarity
measure between images.  There are *many* possibilities and that
measure will be crucial.  I've heard of situations where only
the color distribution histogram was used of each image (and
similarity ~= some kind correlation).  
You should also read a bit about cluster analysis in a
textbook (on "clustering" or "multivariate analysis").
For the cluster package, the obvious textbook is
Kaufman & Rousseeuw (1990).
help(agnes) has a "BACKGROUND" and "References" section.

    Francisco> Tks,

Martin Maechler <maechler at stat.math.ethz.ch>	http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum  LEO C16	Leonhardstr. 27
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1228			<><

More information about the R-help mailing list