Provides simple code for K-means clustering with deciding the right K and scores the new dataset for the right clusters.

K_Means_Clustering.R

#read data in r

iris <- read.csv("C:/Users/Ashwin/Desktop/segmentation/CSV Fishers Iris Data.csv")
View(iris)
summary(iris)
head(iris)

# Randomise data for making little realistic

iris<-iris[sample(1:nrow(iris)),]
View(iris)
summary(iris)
head(iris)

# Plot data for general visualisation

library(ggplot2)
ggplot(iris, aes(Petal_Length, Petal_Width, color = Species)) + geom_point()

# Remove the species variable so that it does not get included as classification var

iris1 <- iris
iris1$Species <- NULL
View(iris1)
iris1_scale<-scale(iris1[,2:ncol(iris1)])
View(iris1_scale)

# Run kmeans clustering

km.out <- kmeans(iris1_scale, centers=3, nstart=20)
summary(km.out)
print(km.out)

# confusion matrix
table(iris$Species,km.out$cluster)


# Other metrics

# Overall R square
Overall_R_sq<-km.out$betweenss/km.out$totss
Overall_R_sq  
  
  

# Plots

plot(iris[c("Petal_Length", "Petal_Width")],col=km.out$cluster
     ,main = "k-means with 3 clusters")
plot(iris[c("Petal_Length", "Petal_Width")],col=iris$Species)
plot(iris[c("Sepal_Length", "Sepal_Width")],col=km.out$cluster)

# Use of nstart

# Set up 2 x 3 plotting grid

par(mfrow = c(2, 3))
# Set seed
set.seed(2)
for(i in 1:6) {
# Run kmeans() on x with three clusters and one start
km.out1 <- kmeans(iris1_scale, centers = 3, nstart = 1)
# Plot clusters
plot(iris1[c("Sepal_Length", "Sepal_Width")], col = km.out1$cluster, main = km.out1$tot.withinss)
}


# Finding optimal number of clusters

# Initialize total within sum of squares error: wss
wss <- 0

# For 1 to 15 cluster centers
for (i in 1:15) {
  km.out2 <- kmeans(iris1_scale, centers = i, nstart = 20)
  # Save total within sum of squares to wss variable
  wss[i] <- km.out2$tot.withinss
}

par(mfrow = c(1, 1))

# Plot total within sum of squares vs. number of clusters
plot(1:15, wss, type = "b", 
     xlab = "Number of Clusters", 
     ylab = "Within groups sum of squares")



# ------------------------------------------------------------------


# Apply the clustering to new data

iris_test <- read.csv("C:/Users/Ashwin/Desktop/segmentation/Fishers Iris Data -for scoring.csv")
head(iris_test)

# Randomise data for making little realistic

iris_test<-iris_test[sample(1:nrow(iris)),]
View(iris_test)
summary(iris_test)
head(iris_test)

# Plot data for general visualisation

library(ggplot2)
ggplot(iris_test, aes(Petal_Length, Petal_Width, color = Species)) + geom_point()
ggplot(iris, aes(Petal_Length, Petal_Width, color = Species)) + geom_point()

# Remove the species variable so that it does not get included as classification var

iris_test1 <- iris_test
iris_test1$Species <- NULL
View(iris_test1)
iris_test1_scale<-scale(iris1[,2:ncol(iris1)])

# Creating closest cluster function

closest.cluster <- function(x) {
  cluster.dist <- apply(km.out$centers, 1, function(y) sqrt(sum((x-y)^2)))
  return(which.min(cluster.dist)[1])
}


clusters2 <- apply(iris_test1_scale, 1, closest.cluster)
clusters2
clusters2.df<-data.frame(clusters2)
View(clusters2.df)
# Attaching clusters information to original without clusters data
iris_test2<-merge(x=iris_test1,y=clusters2)
View(iris_test2)


install.packages("NbClust")
library(NbClust)
I_ccc <- NbClust(iris1, distance="euclidean", min.nc=2, max.nc=15, method = "kmeans",
                 index = "ccc")
I_ccc