FPDclustering (version 2.3.1)

Silh: Probabilistic silhouette plot

Description

Graphical tool to evaluate the clustering partition.

Usage

Silh(p)

Value

Probabilistic silhouette plot

Arguments

p

A matrix of probabilities such that rows correspond to observations and columns correspond to clusters.

Author

Cristina Tortora

Details

The probabilistic silhouettes are an adaptation of the ones proposed by Menardi(2011) according to the following formula: $$dbs_i = (log(p_{im_k}/p_{im_1}))/max_i |log(p_{im_k}/p_{im_1})|$$

where \(m_k\) is such that \(x_i\) belongs to cluster \(k\) and \(m_1\) is such that \(p_{im_1}\) is maximum for \(m\) different from\(m_k\).

References

Menardi G. Density-based Silhouette diagnostics for clustering methods.Statistics and Computing, 21, 295-308, 2011.

Examples

if (FALSE) {
# Asymmetric data set silhouette example (with shape=3).
data('asymmetric3')
x<-asymmetric3[,-1]
fpdas3=FPDC(x,4,3,3)
Silh(fpdas3$probability)
}

if (FALSE) {
# Asymmetric data set shiluette example (with shape=20).
data('asymmetric20')
x<-asymmetric20[,-1]
fpdas20=FPDC(x,4,3,3)
Silh(fpdas20$probability)
}

if (FALSE) {
# Shiluette example with outliers.
data('outliers')
x<-outliers[,-1]
fpdout=FPDC(x,4,4,3)
Silh(fpdout$probability)
}