Clustering with a Gaussian Mixture Model

In [25]:

    

%matplotlib inline
import matplotlib.pyplot as plt
import mpl_toolkits.mplot3d.axes3d as p3

import pandas as pd
import numpy as np


from sklearn import datasets
from sklearn import mixture
from sklearn import manifold


    

In [26]:

    

data, labels_true = datasets.make_blobs(n_samples=750, centers=[[1,1],[0,5],[2,8]], cluster_std=0.7,
                            random_state=0)


plt.scatter(data[:,0], data[:,1])

df = pd.DataFrame(data, columns=['X', 'Y'])


    

In [27]:

    

gmm = mixture.GaussianMixture(n_components=2)
gmm.fit(data)
label = gmm.predict(data)
df['label'] = label


fig = plt.figure()
fig.suptitle('GaussianMixture n=2', fontsize=14, fontweight='bold')
ax = fig.add_subplot(111)

clusters = list(set(label))

for i in range(len(clusters)):
    plt.scatter(df[df.label == clusters[i]].X, df[df.label == clusters[i]].Y, 
                label=clusters[i], color=plt.cm.jet(np.float(i) / len(np.unique(label))))

    
plt.legend(bbox_to_anchor=(1.25, 1))


    

    
Out[27]:





<matplotlib.legend.Legend at 0x10dab0a58>






    

In [28]:

    

gmm = mixture.GaussianMixture(n_components=3)
gmm.fit(data)
label = gmm.predict(data)
df['label'] = label


fig = plt.figure()
fig.suptitle('GaussianMixture n=3', fontsize=14, fontweight='bold')
ax = fig.add_subplot(111)

clusters = list(set(label))

for i in range(len(clusters)):
    plt.scatter(df[df.label == clusters[i]].X, df[df.label == clusters[i]].Y, 
                label=clusters[i], color=plt.cm.jet(np.float(i) / len(np.unique(label))))

    
plt.legend(bbox_to_anchor=(1.25, 1))


    

    
Out[28]:





<matplotlib.legend.Legend at 0x10d50dba8>






    

In [29]:

    

gmm = mixture.GaussianMixture(n_components=4)
gmm.fit(data)
label = gmm.predict(data)
df['label'] = label


fig = plt.figure()
fig.suptitle('GaussianMixture n=4', fontsize=14, fontweight='bold')
ax = fig.add_subplot(111)

clusters = list(set(label))

for i in range(len(clusters)):
    plt.scatter(df[df.label == clusters[i]].X, df[df.label == clusters[i]].Y, 
                label=clusters[i], color=plt.cm.jet(np.float(i) / len(np.unique(label))))

    
plt.legend(bbox_to_anchor=(1.25, 1))


    

    
Out[29]:





<matplotlib.legend.Legend at 0x10d264198>






    

In [30]:

    

data, t = datasets.make_swiss_roll(n_samples=200, noise=0.1, random_state=0)

df = pd.DataFrame(data, columns=['X', 'Y', 'Z'])


fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)

ax.scatter(df.X, df.Y, df.Z, 'o')


    

    
Out[30]:





<mpl_toolkits.mplot3d.art3d.Path3DCollection at 0x10d5b2588>






    

In [31]:

    

gmm = mixture.GaussianMixture(n_components=2)
gmm.fit(data)
label = gmm.predict(data)

df['label'] = label

fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)
fig.suptitle('GaussianMixture n=2', fontsize=14, fontweight='bold')

for l in np.unique(label):
    
    ax.scatter(df[df.label == l].X, df[df.label == l].Y, df[df.label == l].Z, 
               'o', color=plt.cm.jet(np.float(l) / len(np.unique(label))))


    

In [32]:

    

gmm = mixture.GaussianMixture(n_components=3)
gmm.fit(data)
label = gmm.predict(data)

df['label'] = label

fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)
fig.suptitle('GaussianMixture n=3', fontsize=14, fontweight='bold')

for l in np.unique(label):
    
    ax.scatter(df[df.label == l].X, df[df.label == l].Y, df[df.label == l].Z, 
               'o', color=plt.cm.jet(np.float(l) / len(np.unique(label))))


    

In [33]:

    

gmm = mixture.GaussianMixture(n_components=4)
gmm.fit(data)
label = gmm.predict(data)

df['label'] = label

fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)
fig.suptitle('GaussianMixture n=4', fontsize=14, fontweight='bold')

for l in np.unique(label):
    
    ax.scatter(df[df.label == l].X, df[df.label == l].Y, df[df.label == l].Z, 
               'o', color=plt.cm.jet(np.float(l) / len(np.unique(label))))


    

In [34]:

    

gmm = mixture.GaussianMixture(n_components=5)
gmm.fit(data)
label = gmm.predict(data)

df['label'] = label

fig = plt.figure()
ax = p3.Axes3D(fig)
ax.view_init(7, -80)
fig.suptitle('GaussianMixture n=5', fontsize=14, fontweight='bold')

for l in np.unique(label):
    
    ax.scatter(df[df.label == l].X, df[df.label == l].Y, df[df.label == l].Z, 
               'o', color=plt.cm.jet(np.float(l) / len(np.unique(label))))