CODEWITHSUNDEEP
rcluster-analysishierarchical-clustering
Abdirizak
Abdirizak

Reputation: 100

Plotting output from hclust()

I have a 300x2 matrix of data, i.e. 300 observations of 2 variables. Using the kmeans function in R, I can plot the resulting clusters in the following way:

data <- scale(data)
fit.kmeans <- kmeans(data, 3)
plot(data, col = fit.kmeans$cluster)

This gives a nice 2D plot of the original data, colored by cluster. Is there any simple way of doing the same using the hclust function? Or, alternatively, is there another function which lets me implement different clustering methods and subsequently plot the resulting clusters? Thanks in advance.

Upvotes: 2

Views: 8943

Answers (2)

Samuel
Samuel

Reputation: 3053

A cluster dendrogram is a nice alternative to a scatter plot for hierarchical clustering:

tree <- hclust(d = dist(x = iris[1:50, 1:4], method = "euclidean"))
cl_members <- cutree(tree = tree, k = 3)
plot(x = tree, labels =  row.names(tree), cex = 0.5)
rect.hclust(tree = tree, k = 3, which = 1:3, border = 1:3, cluster = cl_members)

enter image description here

Upvotes: 6

G5W
G5W

Reputation: 37641

Each clustering method reports the clusters in slightly different ways. In general, you will need to look at the structure returned by the clustering function.

But you ask specifically about hclust. To get the clusters from hclust you need to use the cutree function together with the number of clusters you want. Here is an example of using it with the iris data. 

HC = hclust(dist(iris[,1:4]))
plot(iris[,3:4], pch=20, col=cutree(HC,3))

Clusters

Upvotes: 2

Related Questions