clusterclara {goCluster}R Documentation

Clusters a dataset with the clara function

Description

This function is used in the goCluster framework to cluster a dataset with the clara function.

Usage

clusterclara(dataset, clusters, distance = "euclidean", repeats = 1, fixed = TRUE)

Arguments

dataset The dataset to be clustered. This has to be a matrix.
clusters This specifies the number of clusters that the dataset should be partitioned into.
distance The distance metric that is going to be used by clara.
fixed This option determines whether the analysis should start with a fixed random seed or with a truely random seed. A fixed seed leads to a stable result but does not represent the inherent variability of the clustering approach.
repeats In case clara clusters without a fixed seed it may be useful to repeat the clustering in order to get an impression of the variability of the clustering result. This option specifies the number of repeats.

Details

Clara clustering will partition the dataset of the parent object into the number of clusters specified by the user. Clara is very similar to PAM (partitioning around medoids) but has been adapted to large datasets. It has the same advantages as PAM concerning the judgement of quality for the resulting clusters but it lacks a deterministic outcome. Therefore it is significantly faster than PAM. You may request a stable result by using a fixed seed, but this can convey an incorrect impression of the stability of the result. Alternatively clusterclara can repeat the clustering though that will partially defeat the gain in speed over PAM.

Value

A "tree" (list of lists) of clusters. The first level will hold as many list elements as the number of times the clustering has been repeated. Each of these elements holds a number of lists equal to the number of clusters requested .Each of node on this second level hold the unique ids of the genes in the cluster.

Author(s)

Gunnar Wrobel, work@gunnarwrobel.de, http://www.gunnarwrobel.de.

See Also

clusterAlgorithmClara-class clara

Examples

require(cluster)

## Get the benomyl setup
data(benomylsetup)

## Extract a fraction of the dataset
benomyldata <- benomylsetup$data$dataset[1:200,]
benomylids  <- benomylsetup$data$uniqueid[1:200]

## Cluster the dataset
clusterclara(exprs(benomyldata), 4)


[Package goCluster version 1.6.0 Index]