Package 'ockc'

Title: Order Constrained Solutions in k-Means Clustering
Description: Extends 'flexclust' with an R implementation of order constrained solutions in k-means clustering (Steinley and Hubert, 2008, <doi:10.1007/s11336-008-9058-z>).
Authors: Sebastian Krey [aut, cre], Friedrich Leisch [aut], Sebastian Hoffmeister [ctb]
Maintainer: Sebastian Krey <[email protected]>
License: GPL-2 | GPL-3
Version: 1.1
Built: 2024-12-12 06:52:33 UTC
Source: CRAN

Help Index

Bootstrap Order Constrained k-means Clustering

Description

Runs ockc for different numbers of clusters on bootstrap replica of the original data (maintaining the supplied order) and returns corresponding cluster assignments, centroids and Rand indices comparing pairs of partitions.

Usage

bootockc(x, k, nboot = 100, order = NULL, correct = TRUE, seed = NULL,
         multicore = TRUE, verbose = FALSE, ...)

Arguments

x, k, ...

Passed to ockc

nboot

Number of bootstrap pairs (maintaining order).

order

Order restriction of x. If NULL an initial run of ockc with order=NULL is run to calculate an order with seriate from package seriation

correct

Logical, correct the index for agreement by chance?

seed

If not NULL, a call to set.seed() is made before any clustering is done.

multicore

Use parallelization, if available. For examples and additional documentation see bootFlexclust.

verbose

Logical, show progress information during computations. Ignored if multicore=TRUE.

Value

Returns an object of class "bootFlexclust".

Author(s)

Sebastian Krey

See Also

ockc, bootFlexclust, stepFlexclust

Examples

x <- rbind(cbind(rnorm(10, mean=0), rnorm(10, mean=0,), rnorm(10, mean=0)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=10), rnorm(10, mean=0)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=0), rnorm(10, mean=10)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=10), rnorm(10, mean=10))
           )

bockc <- bootockc(x, 2:4, nboot=4, order=c(1:10, 21:40, 11:20),
                  multicore=FALSE, verbose=FALSE)
bockc

Order Contrained Solutions in k-Means Clustering

Description

Calculates an order constrained clustering solution (default k-means) on a data matrix.

Usage

ockc(x, k, family = kccaFamily("kmeans"), order = NULL, control = NULL,
     save.data = FALSE, multicore = FALSE, ...)

Arguments

x

A numeric matrix of data.

k

An integer vector of number of clusters. For each element of k a clustering solution is computed (reusage of intermediate results makes this more efficient than individual calls of ockc).

family

Object of class kccaFamily.

order

Order restriction of x. If NULL an order is caluclated with seriate from package seriation

control

An object of class flexclustControl.

save.data

Save a copy of x in the return object?

multicore

Use parallelization, if available. For examples and additional documentation see bootFlexclust.

...

Additional options for seriate for order calculation.

Author(s)

Sebastian Krey, Friedrich Leisch, Sebastian Hoffmeister

References

Steinley, D. and Hubert, L. (2008). Order-Constrained Solutions in K-Means Clustering: Even Better Than Being Globally Optimal. Psychometrika, 73 (4), pp. 647-664.

See Also

kcca

Examples

x <- rbind(cbind(rnorm(10, mean=0), rnorm(10, mean=0,), rnorm(10, mean=0)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=10), rnorm(10, mean=0)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=0), rnorm(10, mean=10)),
           cbind(rnorm(10, mean=10), rnorm(10, mean=10), rnorm(10, mean=10))
           )

res <- ockc(x, k=4, nboot=4, order=c(1:10, 21:40, 11:20))
res