clustCombiOptim: Optimal number of clusters obtained by combining mixture...

In mclust: Gaussian Mixture Modelling for Model-Based Clustering, Classification, and Density Estimation

Optimal number of clusters obtained by combining mixture components

Description

Return the optimal number of clusters by combining mixture components based on the entropy method discussed in the reference given below.

Usage

clustCombiOptim(object, reg = 2, plot = FALSE, ...)

Arguments

object	An object of class 'clustCombi' resulting from a call to clustCombi.
reg	The number of parts of the piecewise linear regression for the entropy plots. Choose 2 for a two-segment piecewise linear regression model (i.e. 1 change-point), and 3 for a three-segment piecewise linear regression model (i.e. 3 change-points).
plot	Logical, if TRUE an entropy plot is also produced.
...	Further arguments passed to or from other methods.

Value

The function returns a list with the following components:

numClusters.combi	The estimated number of clusters.
z.combi	A matrix whose [i,k]th entry is the probability that observation i in the data belongs to the kth cluster.
cluster.combi	The clustering labels.

Author(s)

J.-P. Baudry, A. E. Raftery, L. Scrucca

References

J.-P. Baudry, A. E. Raftery, G. Celeux, K. Lo and R. Gottardo (2010). Combining mixture components for clustering. Journal of Computational and Graphical Statistics, 19(2):332-353.

See Also

combiPlot, entPlot, clustCombi

Examples

data(Baudry_etal_2010_JCGS_examples)
output <- clustCombi(data = ex4.1) 
combiOptim <- clustCombiOptim(output)
str(combiOptim)

# plot optimal clustering with alpha color transparency proportional to uncertainty
zmax <- apply(combiOptim$z.combi, 1, max)
col <- mclust.options("classPlotColors")[combiOptim$cluster.combi]
vadjustcolor <- Vectorize(adjustcolor)
alphacol = (zmax - 1/combiOptim$numClusters.combi)/(1-1/combiOptim$numClusters.combi)
col <- vadjustcolor(col, alpha.f = alphacol)
plot(ex4.1, col = col, pch = mclust.options("classPlotSymbols")[combiOptim$cluster.combi])

mclust documentation built on Nov. 16, 2023, 5:10 p.m.