Name: Tushar Nankani

Batch: C23
Roll No: 1902112
Question Number: 10

DWM PRACTICAL EXAM

AIM
We have several objects (5 types of medicines) and each object have
two attributes or features Weight Index and as pH shown in table
below.
Cluster the objects using hierarchical clustering.

Medicine WeightIndex pH
A 1 3
B 2 5
C 5 4
D 1 2

E 5 6
Solution
Data Cleaning:
Since an entire row is missing, we will have to ignore the tuple
completely in the data cleaning process.

Cleaned and updated dataset:

Medicine WeightIndex pH
A 1 3
B 2 5
C 5 4
D 1 2
E 5 6

For hierarchical clustering:

• Importing Necessary Libraries

import math
import pandas as pd
import numpy as np
import scipy.cluster.hierarchy as shc
import matplotlib.pyplot as plt
 • Loading the data in the notebook:
data = {
'Medicine' : ['A', 'B', 'C', 'D', 'E'],
'WeightIndex' : [1, 2, 5, 1, 5],
'pH' : [3, 5, 4, 2, 6]
}
data = pd.DataFrame.from_dict(data)

• Final dataset for clustering

data = data[['WeightIndex', 'pH']]
• calculate_dist function between 2 clusters

def calculate_dist(cluster1, cluster2):

maximum_dist = -1
for point1 in cluster1:
for point2 in cluster2:

dist = 0
for i in range(len(point1)-1):
dist += math.pow((point1[i] - point2[i]), 2)

dist = math.sqrt(dist)

if dist > maximum_dist:

maximum_dist = dist

return maximum_dist

• renaming_clusters function to rename the clusters

def renaming_clusters(dataset, num):

clusters = list(pd.unique(dataset['Cluster']))
new_clusters = list(range(num))

mapping = {}
for key, val in zip(clusters, new_clusters):
mapping[key] = val

dataset["Cluster"] = dataset["Cluster"].map(lambda x:
mapping[x])
return dataset
• hierarchical_clustering function between 2 clusters

def hierarchical_clustering(dataset, num):

dataset['Cluster'] = dataset.index

while (True):
clusters = []

for i in range(max(pd.unique(dataset['Cluster']))
+ 1):
cl =
dataset[dataset['Cluster']==i].values.tolist()
if len(cl) > 0:
clusters.append(cl)

if len(clusters) == num:
dataset = renaming_clusters(dataset,num)
return dataset

minimum_dist = math.inf
c1 = -1
c2 = -1

for i in range(len(clusters) - 1):

for j in range(i + 1, len(clusters)):

cluster1 = clusters[i]
cluster2 = clusters[j]
dist = calculate_dist(cluster1, cluster2)

if dist < minimum_dist:

minimum_dist = dist
c1 = cluster1[0][-1]
c2 = cluster2[0][-1]

dataset.loc[dataset.Cluster == max(c1, c2),

"Cluster"] = min(c1, c2)
• Creating a Dendogram for predicting the approximate number of
clusters.

import scipy.cluster.hierarchy as shc

import matplotlib.pyplot as plt

plt.figure(figsize=(10, 7))
plt.title("Dendrograms")
dend = shc.dendrogram(shc.linkage(data, method='ward'))

• The above dendogram shows that the ideal number of

clusters.

num = int(input('Enter number of clusters : '))

result = hierarchical_clustering(data, num)
result
for i in range(num):
plt.scatter(result[result['Cluster'] ==
i][result.columns[0]], result[result['Cluster'] ==
i][result.columns[1]], s = 100, label = 'Cluster
{}'.format(i))

plt.title('{} Clusters'.format(num))
plt.xlabel('Weighted Index')
plt.ylabel('pH')
plt.legend()
plt.show()
FINAL OUTPUT:
10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

Name: Tushar Nankani

Roll No: 1902112

Batch: C23

In [1]: import pandas as pd

import numpy as np
import math

In [2]: data = {
'Medicine' : ['A', 'B', 'C', 'D', 'E'],
'WeightIndex' : [1, 2, 5, 1, 5],
'pH' : [3, 5, 4, 2, 6]
}

In [3]: data = pd.DataFrame.from_dict(data)

In [4]: data

Out[4]:
Medicine WeightIndex pH

0 A 1 3

1 B 2 5

2 C 5 4

3 D 1 2

4 E 5 6

In [8]: data = data[['WeightIndex', 'pH']]

In [9]: data

Out[9]:
WeightIndex pH

0 1 3

1 2 5

2 5 4

3 1 2

4 5 6

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 1/6

10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

In [12]: def calculate_dist(cluster1, cluster2):

maximum_dist = -1
for point1 in cluster1:
for point2 in cluster2:

dist = 0
for i in range(len(point1)-1):
dist += math.pow((point1[i] - point2[i]), 2)
​
dist = math.sqrt(dist)

if dist > maximum_dist:

maximum_dist = dist

return maximum_dist

In [13]: def renaming_clusters(dataset, num):

clusters = list(pd.unique(dataset['Cluster']))
new_clusters = list(range(num))

mapping = {}
for key, val in zip(clusters, new_clusters):
mapping[key] = val

dataset["Cluster"] = dataset["Cluster"].map(lambda x: mapping[x])

return dataset

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 2/6

10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

In [35]: def hierarchical_clustering(dataset, num):

dataset['Cluster'] = dataset.index

while (True):
clusters = []

for i in range(max(pd.unique(dataset['Cluster'])) + 1):

cl = dataset[dataset['Cluster']==i].values.tolist()
if len(cl) > 0:
clusters.append(cl)

if len(clusters) == num:
dataset = renaming_clusters(dataset,num)
return dataset

minimum_dist = math.inf
c1 = -1
c2 = -1
​
for i in range(len(clusters) - 1):
for j in range(i + 1, len(clusters)):

cluster1 = clusters[i]
cluster2 = clusters[j]
dist = calculate_dist(cluster1, cluster2)

if dist < minimum_dist:

minimum_dist = dist
c1 = cluster1[0][-1]
c2 = cluster2[0][-1]

dataset.loc[dataset.Cluster == max(c1, c2), "Cluster"] = min(c1, c2)

In [36]: import scipy.cluster.hierarchy as shc

import matplotlib.pyplot as plt

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 3/6

10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

In [37]: plt.figure(figsize=(10, 7))

plt.title("Dendrograms")
dend = shc.dendrogram(shc.linkage(data, method='ward'))

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 4/6

10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

In [40]: #The above dendogram shows that the ideal number of clusters should be 2
num = int(input('Enter number of clusters : '))
result = hierarchical_clustering(data, num)
result

Enter number of clusters : 3

Out[40]:
WeightIndex pH Cluster

0 1 3 0

1 2 5 1

2 5 4 2

3 1 2 0

4 5 6 2

In [41]: for i in range(num):

plt.scatter(result[result['Cluster'] == i][result.columns[0]], result[result[
​
plt.title('{} Clusters'.format(num))
plt.xlabel('Weighted Index')
plt.ylabel('pH')
plt.legend()
plt.show()

In [ ]: ​

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 5/6

10/30/21, 10:58 AM DWM_Pracs - Jupyter Notebook

localhost:8888/notebooks/Desktop/COLLEGE/SEM 5/DWM_Pracs.ipynb 6/6