---
title: R: K-Means
---
What does K-MEANS do ?
Program K-MEANS computes a non-hierarchical clustering by
minimization of the within-group variance, following several variants
of the method originally proposed by
MacQueen (1967), to which he gave
the name
k-means. This is a
partitioning method for a
group of objects, and not a method of hierarchical classification. The
user decides how many groups,
k, she wants to obtain from the
program. The
k-means algorithm followed here is the one
described on page 163 of
Anderberg (1973).
The present program computes
the clustering with or without contiguity constraint (spatial or
temporal), following the user's request. It complements the
hierarchical clustering programs of the R package, which
implement various algorithms of clustering without constraint (
CLUSTER
in the Macintosh version) or with contiguity constraint (
BIOGEO and
CHRONO).
Last updated on Saturday, March 30, 2013 by Philippe Casgrain