# ---------- K-Means ------------
Data <- read.csv("Data.csv", header = F)

plot(Data)

kData = kmeans(Data, centers = 2)

plot(Data, col=kData$cluster)

# ---------- Hierarchical Clustering ------------
par(mfrow=c(2,2))
plot(hclust(dist(Data), method = "single"), main = "Single Linkage", labels = FALSE)
plot(hclust(dist(Data), method = "complete"), main = "Complete Linkage", labels = FALSE)
plot(hclust(dist(Data), method = "average"), main = "Average Linkage", labels = FALSE)
plot(hclust(dist(Data), method = "centroid"), main = "Centroid Linkage", labels = FALSE)
dev.off()


# ---------- K-Means, Again ------------

max_clusters = 20
sum_dist = vector(length = max_clusters)
for (k in 1:max_clusters) {
  kData = kmeans(Data, centers = k)
  sum_dist[k] = kData$tot.withinss
}

plot(x = seq(1,20), y = sum_dist, col = 'blue',
     xlab = 'num of clusters', ylab = 'sum of within-group distances')
lines(x = seq(1,20), y = sum_dist, col = 'blue')
axis(1,labels = seq(1,20), at = seq(1,20))

# To create a silhouette graph we need the 'cluster' and 'factoextra' packages
if(!("cluster" %in% rownames(installed.packages()))) {install.packages("cluster")}

library(cluster)

par(mfrow=c(3,2))
for(i in 2:6){
  kData = kmeans(Data, centers = i)
  plot(silhouette(kData$cluster, dist(Data)), col = 1:i, border = NA, do.n.k = F, do.clus.stat = T,
       cex.lab=1.5, cex.axis=1.5, cex.sub = 1.5, cex.names = 1.5, main = "")
}
dev.off()

silAvg = matrix(ncol = 2, nrow = 5)
for(i in 2:6){
  kData = kmeans(Data, centers = i)
  silAvg[i-1,1] = i
  silAvg[i-1,2] = mean(silhouette(kData$cluster, dist(Data))[,3])
}

plot(silAvg, xlab = "Number of clusters", ylab = "Average silhouette value")
lines(silAvg)