1 from __future__ import print_function
3 from sklearn.datasets import make_blobs
4 from sklearn.cluster import KMeans
5 from sklearn.metrics import silhouette_samples, silhouette_score
7 import matplotlib.pyplot as plt
8 import matplotlib.cm as cm
13 # Generating the sample data from make_blobs
14 # This particular setting has one distinct cluster and 3 clusters placed close
16 '''X, y = make_blobs(n_samples=500,
20 center_box=(-10.0, 10.0),
22 random_state=1) # For reproducibility'''
24 X = np.array([[132, 192], [117, 960], [117, 962], [1343, 0], [117, 1116], [117, 1117], [117, 1118], [117, 1119], [1015, 0], [117, 966]])
26 range_n_clusters = [2, 3, 4, 5, 6]
28 for n_clusters in range_n_clusters:
29 # Create a subplot with 1 row and 2 columns
30 fig, (ax1, ax2) = plt.subplots(1, 2)
31 fig.set_size_inches(18, 7)
33 # The 1st subplot is the silhouette plot
34 # The silhouette coefficient can range from -1, 1 but in this example all
35 # lie within [-0.1, 1]
36 ax1.set_xlim([-0.1, 1])
37 # The (n_clusters+1)*10 is for inserting blank space between silhouette
38 # plots of individual clusters, to demarcate them clearly.
39 ax1.set_ylim([0, len(X) + (n_clusters + 1) * 10])
41 # Initialize the clusterer with n_clusters value and a random generator
42 # seed of 10 for reproducibility.
43 clusterer = KMeans(n_clusters=n_clusters, random_state=10)
44 cluster_labels = clusterer.fit_predict(X)
46 # The silhouette_score gives the average value for all the samples.
47 # This gives a perspective into the density and separation of the formed
49 silhouette_avg = silhouette_score(X, cluster_labels)
50 print("For n_clusters =", n_clusters,
51 "The average silhouette_score is :", silhouette_avg)
53 # Compute the silhouette scores for each sample
54 sample_silhouette_values = silhouette_samples(X, cluster_labels)
57 for i in range(n_clusters):
58 # Aggregate the silhouette scores for samples belonging to
59 # cluster i, and sort them
60 ith_cluster_silhouette_values = \
61 sample_silhouette_values[cluster_labels == i]
63 ith_cluster_silhouette_values.sort()
65 size_cluster_i = ith_cluster_silhouette_values.shape[0]
66 y_upper = y_lower + size_cluster_i
68 color = cm.nipy_spectral(float(i) / n_clusters)
69 ax1.fill_betweenx(np.arange(y_lower, y_upper),
70 0, ith_cluster_silhouette_values,
71 facecolor=color, edgecolor=color, alpha=0.7)
73 # Label the silhouette plots with their cluster numbers at the middle
74 ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
76 # Compute the new y_lower for next plot
77 y_lower = y_upper + 10 # 10 for the 0 samples
79 ax1.set_title("The silhouette plot for the various clusters.")
80 ax1.set_xlabel("The silhouette coefficient values")
81 ax1.set_ylabel("Cluster label")
83 # The vertical line for average silhouette score of all the values
84 ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
86 ax1.set_yticks([]) # Clear the yaxis labels / ticks
87 ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
89 # 2nd Plot showing the actual clusters formed
90 colors = cm.nipy_spectral(cluster_labels.astype(float) / n_clusters)
91 ax2.scatter(X[:, 0], X[:, 1], marker='.', s=30, lw=0, alpha=0.7,
92 c=colors, edgecolor='k')
94 # Labeling the clusters
95 centers = clusterer.cluster_centers_
96 # Draw white circles at cluster centers
97 ax2.scatter(centers[:, 0], centers[:, 1], marker='o',
98 c="white", alpha=1, s=200, edgecolor='k')
100 for i, c in enumerate(centers):
101 ax2.scatter(c[0], c[1], marker='$%d$' % i, alpha=1,
104 ax2.set_title("The visualization of the clustered data.")
105 ax2.set_xlabel("Feature space for the 1st feature")
106 ax2.set_ylabel("Feature space for the 2nd feature")
108 plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "
109 "with n_clusters = %d" % n_clusters),
110 fontsize=14, fontweight='bold')