CLARA: CLARA clustering

In talegari/clusterfit: An interface for clustering in R

Description

Implements CLARA clustering algorithm using pam

Usage

CLARA(x, k, nSamples = 5, sampleFrac = 0.1, swap = FALSE,
  pamonce = 0)

Arguments

x	(numeric matrix or dist) data
k	(positive integer) Number of clusters
nSamples	(positive integer, default: 5) Number of random samples
sampleFrac	(positive fraction, default: 0.1) Fraction of observations in a sample
swap	(flag, default: FALSE) Whether PAM should involve swap phase
pamonce	(One among 0, 1, 2, default: 0) See pamonce argument in pam

Details

CLARA implementation:

• PAM clustering is computed on multiple random samples of observations.

• For a given clustering/medoids, cost is defined as the average dissimilarity/distance between observations(entire dataset) from the nearest medoid.

• A clustering/medoids corresponding to the clustering with minimum cost is chosen.

The PAM fitting on multiple subsets is parallelized with future.

Value

A list with three compoments:

• clustering: An integer vector indicating the cluster number with length equal to number of observations

• medoidsIndex: An integer vector of indices of medoids

• cost: average dissimilarity/distance between observations(entire dataset) from the nearest medoid

Author(s)

Srikanth Komala Sheshachala ([e-mail])

Examples

set.seed(1)
clara(dist(mtcars), k = 4, sampleFrac = 0.4, nSamples = 10)
set.seed(2)
clara(stats::dist(mtcars, method = "maximum"), k = 4, sampleFrac = 0.4, nSamples = 10)

talegari/clusterfit documentation built on May 15, 2019, 6:29 p.m.