optimcluster: Optimal clustering via PAM

In lixiangchun/lxctk: Li Xiangchun's tool-kit (lxctk)

Description

Perform optimal clustering via PAM (package cluster) and maximum Silhouette index.

Usage

optimcluster(x, D=dist(x), plot=FALSE)

Arguments

x	Input data, either a matrix, data frame or vector.
D	Distance among entities in x.
plot	Whether to plot clustering metrics.

Value

optim.pam	Optimal pam object.
optim.nclust	Optimal cluster number.
silhouette.index	Silhouette index
Calinski.Harabasz.index	Calinski-Harabasz(CI) index

See Also

pam

Examples

library(cluster)
library(plotrix)
library(MASS)

x=iris[,1:4]
D=dist(x)
#pdf('clustering_metrics.pdf', height=5, width=6)
r=optimcluster(x, plot=TRUE)
#dev.off()

# Classical multidimensional scaling (CMDS) of a data matrix, or you can choose
#+to use non-metric multidimensional scaling (NMDS) provided in MASS package
mds=cmdscale(D, 2, eig=TRUE)                  # CMDS
#mds=isoMDS(D, k=2, maxit=5000, tol=1e-6)     # NMDS

# visualize MDS output
#pdf('clustering_mds_figure.pdf', height=5, width=6)
s.class2(mds$points, as.factor(r$optim.pam$clustering), grid=FALSE, clabel=0.5, cpoint=0.9, ellipse.lwd=1)
#dev.off()

lixiangchun/lxctk documentation built on May 21, 2019, 6:44 a.m.