Silhouette widths and Dunn index

SilhouetteWidths: \((b_i - a_i)/\max(a_i, b_i)\) (Rousseeuw 1987), singletons 0, as cluster::silhouette. DunnIndex: smallest between-cluster distance over the largest diameter (Dunn 1974).

Usage

SilhouetteWidths(D, clusters)

DunnIndex(D, clusters)

Arguments

D

Dissimilarity matrix.

clusters

Labels.

Value

List or number.

References

Rousseeuw, P. J. (1987). Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics 20, 53-65.

Examples

D <- rbind(c(0, 1, 5, 6), c(1, 0, 4, 5), c(5, 4, 0, 1), c(6, 5, 1, 0))
SilhouetteWidths(D, c(1, 1, 2, 2))$width
#> [1] 0.8181818 0.7777778 0.7777778 0.8181818
DunnIndex(D, c(1, 1, 2, 2))
#> [1] 4