### Different approaches for clustering
### CSE 8331 - Spring 2012 - Michael Hahsler


### create data
library("mlbench")
set.seed(1234)
Cassini <- mlbench.cassini(1000)
plot(Cassini)
x<-Cassini$x
k<-3

### k-means
cl <- kmeans(x, centers=k)
plot(x, col=cl$cluster+1)
points(cl$centers, cex=5, lwd=2)

### k-medoits
library("cluster")
cl <- pam(x, k=k)
plot(x, col=cl$cluster+1)
points(cl$medoids, cex=5, lwd=2)


### hierarchical clustering
d<- dist(x)

hc <- hclust(d) ### default is complete linkage
plot(hc)
cl <- cutree(hc, k=k)
plot(x, col=cl+1)

hc <- hclust(d, method="single")
plot(hc)
cl <- cutree(hc, k=k)
plot(x, col=cl+1)

hc <- hclust(d, method="average")
plot(hc)
cl <- cutree(hc, k=k)
plot(x, col=cl+1)

### density based clustering
library(fpc)
cl <- dbscan(x, eps=.2, MinPts=4)
plot(x, col=cl$cluster+1)


### model based clustering (estimates a Gaussian mixture model using EM and BIC)
library(mclust)
cl <- Mclust(x, G=1:(k+1))   ### try 1 to 6 components
plot(x, col=cl$classification+1)


### try other dataset
twonorm <- mlbench.twonorm(500, d=2)
plot(twonorm)
x <- twonorm$x
k <- 2

Smiley <- mlbench.smiley()
plot(Smiley)
x <- Smiley$x
k <- 4

Spirals <- mlbench.spirals(500,1,0.05)
plot(Spirals)
x <- Spirals$x
k <- 2