-
Notifications
You must be signed in to change notification settings - Fork 0
/
hierarchical.py
120 lines (105 loc) · 4.8 KB
/
hierarchical.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
""" Agglomerative (Bottom-Up) Clustering Algorithm"""
import numpy as np
np.random.seed(0)
import pandas as pd
import matplotlib.pyplot as plt
from scipy.spatial.distance import pdist
from scipy.cluster import hierarchy
from kmeans import compute_wcc, compute_sc, compute_nmi, plot_clusters
colors = ['blue', 'green', 'red', 'cyan', 'magenta', 'yellow', 'black', 'purple', 'pink', 'brown', \
'orange', 'teal', 'coral', 'lightblue', 'lime', 'lavender', 'turquoise', 'darkgreen', 'tan', 'salmon', \
'gold', 'lightpurple', 'darkred', 'darkblue']
def hierarchical(data, linkage):
ytdist = pdist(data)
Z = hierarchy.linkage(ytdist, linkage)
return Z
def plot_dendrograms(Z1, Z2, Z3):
_, axes = plt.subplots(3, 1, figsize=(15, 8))
axes[0].set_title('Dendrogram using Single Linkage')
axes[1].set_title('Dendrogram using Complete Linkage')
axes[2].set_title('Dendrogram using Average Linkage')
axes[0].set_ylabel('Distances')
axes[1].set_ylabel('Distances')
axes[2].set_ylabel('Distances')
dn1 = hierarchy.dendrogram(Z1, ax=axes[0])
dn2 = hierarchy.dendrogram(Z2, ax=axes[1])
dn3 = hierarchy.dendrogram(Z3, ax=axes[2])
plt.tight_layout()
plt.show()
def sample_dataset(df):
new_df = pd.DataFrame()
for idx in range(10):
new_df = new_df.append(df[df[1] == idx].sample(n=10), ignore_index=True)
return new_df
def compute_and_plot_metrics(data, cutree_dataset, n_clusters, linkage, plot=True):
wcc_array = []
sc_array = []
if isinstance(n_clusters, int):
clusters_ = cutree_dataset[:, 0]
centroids_ = []
for k in range(n_clusters):
centroids_.append(np.mean(data[clusters_ == k], axis = 0))
else:
for idx, K in enumerate(n_clusters):
clusters_ = cutree_dataset[:, idx]
centroids_ = []
for k in range(K):
centroids_.append(np.mean(data[clusters_ == k], axis = 0))
wcc = compute_wcc(data, np.array(centroids_), clusters_, K)
sc = compute_sc(data, clusters_, K)
wcc_array.append(wcc)
sc_array.append(sc)
# plot the metrics
if plot:
_, ax = plt.subplots(1,2, figsize=[10,5])
ax[0].plot(n_clusters, wcc_array)
ax[0].set_title('WC SSD vs Number of Clusters (K)')
ax[0].set_xlabel('Number of Clusters (K)')
ax[0].set_ylabel('Within-Cluster Sum Squared Distance')
ax[0].set_xticks(n_clusters)
ax[0].grid(True, linestyle='--', linewidth=0.5)
ax[1].plot(n_clusters, sc_array)
ax[1].set_title('SC vs Number of Clusters (K)')
ax[1].set_xlabel('Number of Clusters (K)')
ax[1].set_ylabel('Silhoutte Coefficient')
ax[1].set_xticks(n_clusters)
ax[1].grid(True, linestyle='--', linewidth=0.5)
plt.suptitle(f'Hierarchical Clustering using {linkage} linkage.')
plt.show()
return clusters_, centroids_
def plot_clusters(linkage, data, centroids, clusters, K):
plt.figure(figsize=(10,8))
plt.title(f'MNIST Dataset Hierarchical Clusters, Linkage {linkage}')
area = (20)**2
for idx in range(K):
plt.scatter(data[clusters == idx][:,0], data[clusters == idx][:,1], c=colors[idx])
plt.scatter(centroids[idx][0], centroids[idx][1], s=area, marker='^', edgecolors='white', c=colors[idx])
plt.show()
def analysis_hierarchical_nmi(data, distances, class_labels, K, linkage):
cutree_dataset = hierarchy.cut_tree(distances, n_clusters=K)
clusters, centroids = compute_and_plot_metrics(data, cutree_dataset, K, 'single', plot=False)
nmi = compute_nmi(class_labels, clusters, K)
print(f'Hierarchical Clustering NMI: {round(nmi,4)} Linkage: {linkage}')
plot_clusters(linkage, data, centroids, clusters, K)
if __name__ == '__main__':
filename = 'digits-embedding.csv'
digits_df = pd.read_csv(filename, header = None)
sampled_data = sample_dataset(digits_df)
data = sampled_data.iloc[:,2:].to_numpy()
class_labels = sampled_data.iloc[:, 1].to_numpy()
Z1 = hierarchical(data, 'single')
Z2 = hierarchical(data, 'complete')
Z3 = hierarchical(data, 'average')
plot_dendrograms(Z1, Z2, Z3)
# 3.4
n_clusters = [2,4,8,16,32]
cutree_dataset1 = hierarchy.cut_tree(Z1, n_clusters=n_clusters)
cutree_dataset2 = hierarchy.cut_tree(Z2, n_clusters=n_clusters)
cutree_dataset3 = hierarchy.cut_tree(Z3, n_clusters=n_clusters)
compute_and_plot_metrics(data, cutree_dataset1, n_clusters, 'single')
compute_and_plot_metrics(data, cutree_dataset2, n_clusters, 'complete')
compute_and_plot_metrics(data, cutree_dataset3, n_clusters, 'average')
# 3.5
analysis_hierarchical_nmi(data, Z1, class_labels, 8, 'single')
analysis_hierarchical_nmi(data, Z2, class_labels, 8, 'complete')
analysis_hierarchical_nmi(data, Z3, class_labels, 8, 'average')