[R] Hierarchical Clustering.ipynb

Kernel: R

Hierarchical Clustering

Data Preprocessing

dataset = read.csv('Mall_Customers.csv')
X = dataset[4:5]

head(X, 10)

Using the dendrogram to find the optimal number of clusters

dendrogram = hclust(dist(X, method = 'euclidean'), method = 'ward.D')
# ward is used to minimize the variance within each clusters
plot(dendrogram,
     main = paste('Dendrogram'),
     xlab = 'Customer',
     ylab = 'Euclidian Distance')

---

From the above dendrogram we can see that the maximum length verticle line whcih does not have any horizontal line is the one providing the number of clusters as 5.

---

Fitting hierarchical clustering to the Mall datset

hc = hclust(dist(X, method = 'euclidean'), method = 'ward.D')
y_hc = cutree(hc, 5)

head(y_hc, 10)

Visualising the clusters

library(cluster)
clusplot(X,
         y_hc,
         lines = 0,
         shade = TRUE,
         color = TRUE,
         labels = 5,
         plotchar = FALSE,
         span = TRUE,
         main = paste('Clusters of customers'),
         xlab = 'Annual Income',
         ylab = 'Spending Score')
# for more see help(clusplot.default)

---

The target customers should be the one with High Earning and High Spend. Here the datapoints inside cluster 1 are the customers with High Earning and High Spend

---