-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
180 lines (134 loc) · 5.37 KB
/
main.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
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
from sklearn.datasets import make_circles
import matplotlib.pyplot as plt
from distance import euclidean
import numpy as np
from builtins import len
import networkx as nx
from load import load_data
import kmeansECT.ECT_KMeans as ECT_KMeans
from sklearn import metrics
from sklearn.cluster import KMeans,AgglomerativeClustering
DATASETS = ['iris','ecoli','glass','pima','sonar','wine','soybean','ionosphere','balance','breast']
DATASET_NAME = 'ecoli'
## select which kind of data you want to use. randomly generated data or one of the UCI datasets
random_data = False
if random_data:
n_samples = 200
k = 2
samples, labels = make_circles(n_samples=n_samples, factor=.3, noise=.05)
bluecircle = samples[labels == 0]
redcircle = samples[labels == 1]
all_circles = np.concatenate([bluecircle, redcircle])
plt.figure()
plt.scatter(bluecircle[:, 0], bluecircle[:, 1], c='b', marker='o', s=10)
plt.scatter(redcircle[:, 0], redcircle[:, 1], c='r', marker='+', s=30)
plt.show()
if random_data == False:
samples, labels, k = load_data.load_dataset(DATASET_NAME)
n_samples = len(samples)
nn = np.sqrt(n_samples)
nn = int(nn)
lst = []
distance_matrix = np.zeros((n_samples, n_samples))
## compute euclidean distance between each pair of data
if random_data:
for i in range(n_samples):
for i in range(n_samples):
for j in range(n_samples):
distance_matrix[i, j] = euclidean.eclidean_distance(all_circles[i], all_circles[j])
else:
for i in range(n_samples):
for j in range(n_samples):
distance_matrix[i, j] = np.linalg.norm(samples[i] - samples[j])
distance_matrix_copy = distance_matrix.copy()
for i in range(n_samples):
temp = distance_matrix_copy[i]
sortedList = sorted(temp)
max_distance = sortedList[nn]
for j in range(n_samples):
if distance_matrix_copy[i, j] > max_distance:
distance_matrix[i, j] = 0
else:
if i!=j and distance_matrix_copy[i, j] != 0:
distance_matrix[i, j] = 1/distance_matrix_copy[i, j]
else:
distance_matrix[i, j] = 0
## convert distance matrix to undirected graph
G = nx.Graph(distance_matrix_copy)
## MST undirected graph
mst_G = nx.minimum_spanning_tree(G)
## convert MST to matrix
MST_matrix = nx.to_numpy_matrix(mst_G)
## convert distance matrix to directed graph
DG = nx.DiGraph(distance_matrix)
## MST directed graph
mst_dg = nx.algorithms.tree.branchings.Edmonds(DG).find_optimum()
## convert MST_D graph to matrix
MST_matrix_D = nx.to_numpy_matrix(mst_dg)
MST = MST_matrix
for i in range(n_samples):
for j in range(n_samples):
if i != j:
if distance_matrix[i,j] == 0.0 :
distance_matrix[i,j] = MST[i,j]
distance_matrix_added_MST = distance_matrix
directed_graph = nx.DiGraph(distance_matrix_added_MST)
markov_chain_matrix = np.zeros((n_samples, n_samples))
ai = np.zeros(n_samples)
for i in range(n_samples):
a_i = np.sum(distance_matrix_added_MST[i])
ai[i] = a_i
for j in range(n_samples):
markov_chain_matrix[i, j] = distance_matrix_added_MST[i, j] / a_i
A = markov_chain_matrix # adjacency matrix
D = np.diag(ai)
L = D - A
V_G = np.sum(ai)
L_Plus = np.linalg.pinv(L)
ECT = np.zeros((n_samples, n_samples))
for i in range(n_samples):
e_i = np.zeros(n_samples)
e_i[i] = 1
for j in range(n_samples):
e_j = np.zeros(n_samples)
e_j[j] = 1
E = e_i - e_j
E = E.reshape(n_samples, 1)
ECT[i, j] = np.sqrt(np.dot(np.dot(np.transpose(E), L_Plus), E) * V_G)
NMI_list_ect_kmeans = {}
NMI_list_kmeans = {}
NMI_list_hierarcy = {}
NUM_ITERATION = 5
sum_nmi_ect_kmeans = 0
sum_nmi_kmeans = 0
sum_nmi_hierarcy = 0
for i in range(NUM_ITERATION):
print("***************************************************iteration :", i)
'''---------------- ECT kmeans ---------------'''
ect_kmeans = ECT_KMeans.ECT_KMeans(samples, k, ECT)
clusters = ect_kmeans.fit()
predict_labels_kmeans = ect_kmeans.labels
NMI_kmeans = metrics.normalized_mutual_info_score(labels, predict_labels_kmeans,average_method='arithmetic')
sum_nmi_ect_kmeans += NMI_kmeans
'''---------------- classic kmeans ---------------'''
k_means = KMeans(n_clusters=k).fit(samples)
predict_labels_kmeans = k_means.labels_
NMI_kmeans = metrics.normalized_mutual_info_score(labels, predict_labels_kmeans)
sum_nmi_kmeans += NMI_kmeans
'''---------------- Hierarchical ----------------'''
hierarchical_clus = AgglomerativeClustering(n_clusters=k)
hierarchical_clus.fit(samples)
predict_labels_hierarchy = hierarchical_clus.labels_
NMI_hierarchy = metrics.normalized_mutual_info_score(labels, predict_labels_hierarchy)
sum_nmi_hierarcy += NMI_hierarchy
NMI_list_ect_kmeans[0] = sum_nmi_ect_kmeans/NUM_ITERATION
NMI_list_kmeans[0] = sum_nmi_kmeans/NUM_ITERATION
NMI_list_hierarcy[0] = sum_nmi_hierarcy/NUM_ITERATION
objects = ('ECT kmeans', 'Classic kmeans', 'Hierarchical')
y_pos = np.arange(len(objects))
performance = [NMI_list_ect_kmeans[0],NMI_list_kmeans[0],NMI_list_hierarcy[0]]
plt.bar(y_pos, performance, align='center', alpha=0.5)
plt.xticks(y_pos, objects)
plt.ylabel('NMI')
plt.title('Clustering Method')
plt.show()