ev.vanDongen: Computes the van Dongen criterion

In lejean2000/CLAV: An R Package For Cluster Validation

Description

The van Dongen criterion is an external validation measure of the representativeness of the majority objects in each class and each cluster. It is an number between 0 and 1 and lower values indicate better cluster similarity. Even though the van Dongen measure is normalized, this is achieved by deleting by 2*n which is an absolute upper bound that is not tight. Hence, a special normalization with a tighter upper bound is also provided (by using normalized=T).

Usage

ev.vanDongen(x, y, normalized = F)

Arguments

x	A vector with cluster assignments.
y	A vector with cluster assignments.
normalized	T or F(default). Indicate whether to normalize the measure or not.

Value

A number between 0 and 1.

References

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.26.9783&rep=rep1&type=pdf

Examples

d<-vegan::vegdist (iris[,3:4], method = "euclidean")
c<-cluster::pam (d, 3, diss = TRUE)
ev.vanDongen(c$clustering, unclass(iris$Species))

lejean2000/CLAV documentation built on Feb. 5, 2020, 8:55 p.m.