factoextra (version 1.0.3)

fviz_silhouette: Visualize Silhouette Information from Clustering

Description

Silhouette (Si) analysis is a cluster validation approach that measures how well an observation is clustered and it estimates the average distance between clusters. fviz_silhouette() provides ggplot2-based elegant visualization of silhouette information from i) the result of silhouette(), pam(), clara() and fanny() [in cluster package]; ii) eclust() and hcut() [in factoextra]. Read more: Clustering Validation Statistics.

Usage

fviz_silhouette(sil.obj, label = FALSE, print.summary = TRUE)

Arguments

sil.obj

an object of class silhouette: pam, clara, fanny [in cluster package]; eclust and hcut [in factoextra].

label

logical value. If true, x axis tick labels are shown

print.summary

logical value. If true a summary of cluster silhouettes are printed in fviz_silhouette().

Value

return a ggplot

Details

- Observations with a large silhouhette Si (almost 1) are very well clustered. - A small Si (around 0) means that the observation lies between two clusters. - Observations with a negative Si are probably placed in the wrong cluster.

See Also

fviz_cluster, hcut, hkmeans, eclust, fviz_dend

Examples

set.seed(123)

# Data preparation
# +++++++++++++++
data("iris")
head(iris)
# Remove species column (5) and scale the data
iris.scaled <- scale(iris[, -5])

# K-means clustering
# +++++++++++++++++++++
km.res <- kmeans(iris.scaled, 3, nstart = 25)

# Visualize kmeans clustering
fviz_cluster(km.res, iris[, -5], frame.type = "norm")+
theme_minimal()

# Visualize silhouhette information
require("cluster")
sil <- silhouette(km.res$cluster, dist(iris.scaled))
fviz_silhouette(sil)

# Identify observation with negative silhouette
neg_sil_index <- which(sil[, "sil_width"] < 0)
sil[neg_sil_index, , drop = FALSE]

# PAM clustering
# ++++++++++++++++++++
require(cluster)
pam.res <- pam(iris.scaled, 3)
# Visualize pam clustering
fviz_cluster(pam.res, frame.type = "norm")+
theme_minimal()
# Visualize silhouhette information
fviz_silhouette(pam.res)

# Hierarchical clustering
# ++++++++++++++++++++++++
# Use hcut() which compute hclust and cut the tree
hc.cut <- hcut(iris.scaled, k = 3, hc_method = "complete")
# Visualize dendrogram
fviz_dend(hc.cut, show_labels = FALSE, rect = TRUE)
# Visualize silhouhette information
fviz_silhouette(hc.cut)