14/09/2018 Tutorial 2 - Clustering

In [13]:

import numpy as np
import pandas as pd
from sklearn.cluster import KMeans
pd.set_option('display.float_format', lambda x: '%.3f' % x)
%matplotlib inline
import matplotlib.pyplot as plt

In [9]:

data = pd.read_csv("./driver_dataset.csv", sep='\t')

In [10]:

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4000 entries, 0 to 3999
Data columns (total 3 columns):
Driver_ID 4000 non-null int64
Distance_Feature 4000 non-null float64
Speeding_Feature 4000 non-null float64
dtypes: float64(2), int64(1)
memory usage: 93.8 KB

In [11]:

data.describe()

Out[11]:

Driver_ID Distance_Feature Speeding_Feature

count 4000.000 4000.000 4000.000

mean 3423312447.500 76.042 10.721

std 1154.845 53.470 13.709

min 3423310448.000 15.520 0.000

25% 3423311447.750 45.248 4.000

50% 3423312447.500 53.330 6.000

75% 3423313447.250 65.632 9.000

max 3423314447.000 244.790 100.000

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 1/7
14/09/2018 Tutorial 2 - Clustering

In [26]:

plt.scatter(data.iloc[:,1:2], data.iloc[:,2:3])
plt.xlabel(data.columns.values[1])
plt.ylabel(data.columns.values[2])
plt.show()

In [28]:

wcss = []
for i in range(1,11):
kmeans = KMeans(n_clusters = i,init = 'k-means++',random_state = 0)
kmeans.fit(data)
wcss.append(kmeans.inertia_)
plt.plot(range(1,11),wcss)
plt.title('The Elbow Method')
plt.xlabel('Number of cluster')
plt.ylabel('WCSS')
plt.show()

In [52]:

kmeans = KMeans(n_clusters = 4,init = 'k-means++',random_state =0)

y_kmeans = kmeans.fit_predict(data)

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 2/7
14/09/2018 Tutorial 2 - Clustering

In [53]:

%matplotlib inline
plt.figsize=(40, 40)
plt.scatter(data.iloc[:,1],data.iloc[:,2], c=y_kmeans)

Out[53]:

<matplotlib.collections.PathCollection at 0x7f381ee64ba8>

In [47]:

from sklearn import preprocessing

#Performing Min_Max Normalization
min_max_scaler = preprocessing.MinMaxScaler()
np_scaled = min_max_scaler.fit_transform(data.iloc[:,1:])
dataN = pd.DataFrame(np_scaled)
dataN.head()

Out[47]:

0 1

0 0.243 0.280

1 0.161 0.250

2 0.214 0.270

3 0.175 0.220

4 0.170 0.250

In [50]:

kmeans = KMeans(n_clusters = 4,init = 'k-means++',random_state =0)

y2_kmeans = kmeans.fit_predict(dataN)

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 3/7
14/09/2018 Tutorial 2 - Clustering

In [59]:

%matplotlib inline
plt.scatter(data.iloc[:,1],data.iloc[:,2], c=y2_kmeans)

Out[59]:

<matplotlib.collections.PathCollection at 0x7f381c32eda0>

In [ ]:

#DBSCAN STARTS

In [78]:

from sklearn.cluster import DBSCAN

dbscan = DBSCAN(eps=0.1, metric='euclidean', min_samples=5)

In [79]:

dbsc = dbscan.fit(data)
dbsc.labels_

Out[79]:

array([-1, -1, -1, ..., -1, -1, -1])

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 4/7
14/09/2018 Tutorial 2 - Clustering

In [80]:

plt.scatter(data.iloc[:,1],data.iloc[:,2], c=dbsc.labels_)

Out[80]:

<matplotlib.collections.PathCollection at 0x7f38142e7550>

In [81]:

dbsc = dbscan.fit(dataN)
dbsc.labels_

Out[81]:

array([0, 0, 0, ..., 1, 1, 1])

In [82]:

plt.scatter(data.iloc[:,1],data.iloc[:,2], c=dbsc.labels_)

Out[82]:

<matplotlib.collections.PathCollection at 0x7f381437b198>

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 5/7
14/09/2018 Tutorial 2 - Clustering

In [66]:

model.labels_

Out[66]:

array([-1, -1, -1, ..., -1, -1, -1])

In [ ]:

#AGGLOMERATIVE STARTS

In [67]:

from sklearn.cluster import AgglomerativeClustering as AC

aggclus = AC(n_clusters = 4,affinity='euclidean',linkage='ward',compute_full_tree='
y_aggclus= aggclus.fit_predict(data.iloc[:,1:3])

In [68]:

y_aggclus

Out[68]:

array([3, 3, 3, ..., 1, 1, 1])

In [69]:

from scipy.cluster.hierarchy import dendrogram, linkage,cut_tree

from scipy.cluster.hierarchy import fcluster
k=4
linkage_matrix = linkage(dataN, "ward",metric="euclidean")
ddata=dendrogram(linkage_matrix,color_threshold=1.5)

In [83]:

ddata=dendrogram(linkage_matrix,color_threshold=1.5)
plt.figure(figsize=(5,7))

Out[83]:

<Figure size 360x504 with 0 Axes>

<Figure size 360x504 with 0 Axes>

http://localhost:8888/notebooks/Documents/BITS%20Course/DM/Tut/TUT2/Piyush_TUT/Tutorial%202%20-%20Clustering.ipynb# 6/7