-
Notifications
You must be signed in to change notification settings - Fork 29
/
utils.py
275 lines (240 loc) · 11.1 KB
/
utils.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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
import numpy as np
import scipy.sparse as sp
import torch
import joblib as jlb
import logging, sys
#logger
logger = logging.getLogger()
logger.setLevel(logging.INFO)
if not logger.handlers:
ch = logging.StreamHandler(sys.stdout)
ch.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch.setFormatter(formatter)
logger.addHandler(ch)
def load_vocab_len(dataset="twitter15"):
with open('dataset/{}/{}_vocab.txt'.format(dataset, dataset), 'rb') as f:
lines = f.readlines()
vocab_size = len(lines)
return vocab_size
def load_data(dataset="twitter15"):
"""
Load citation network dataset (twitter15, twitter16, and weibo)
Loads input corpus from gcn/data directory
ind.dataset.features => the feature vectors of the nodes (tweets and words) as scipy.sparse.csr.csr_matrix object;
ind.dataset.train => the indices of training tweets in nodes;
ind.dataset.dev => the indices of dev tweets in nodes;
ind.dataset.test => the indices of test tweets in nodes;
ind.dataset.labels => the one-hot labels of the all nodes as numpy.ndarray object;
ind.dataset.adj => adjacency matrix of words/tweets nodes as scipy.sparse.csr.csr_matrix object;
All objects above must be saved using python pickle module.
:param dataset: Dataset name
:return: All data input files loaded (as well the training/test data).
"""
print('Loading {} dataset...'.format(dataset))
names = ['features_index', 'train', 'dev', 'test', 'labels', 'adj']
objects = []
for i in range(len(names)):
with open('dataset/{}/ind.{}.{}'.format(dataset, dataset, names[i]), 'rb') as f:
objects.append(jlb.load(f))
# if sys.version_info > (3, 0):
# objects.append(jlb.load(f, encoding='latin1'))
# else:
# objects.append(jlb.load(f))
features_index, train_ids, dev_ids, test_ids, labels, adj = tuple(objects)
# logger.info('features.shape:{}, the length of train_ids:{}, the length of dev_ids:{}, the length of test_ids:{}, labels.shape:{}'.format(features.shape,len(train_ids),len(dev_ids),len(test_ids),labels.shape))
logger.info('features_index.shape:{}, the length of train_ids:{}, the length of dev_ids:{}, the length of test_ids:{}, labels.shape:{}'.format(len(features_index),len(train_ids),len(dev_ids),len(test_ids),labels.shape))
# build symmetric adjacency matrix
adj = adj + adj.T.multiply(adj.T > adj) - adj.multiply(adj.T > adj)
# features = normalize(features.tolil())
# features = features.tolil()
adj = normalize_adj(adj + sp.eye(adj.shape[0]))
# features = torch.FloatTensor(np.array(features.todense()))
labels = torch.LongTensor(labels)
adj = sparse_mx_to_torch_sparse_tensor(adj)
idx_train = torch.LongTensor(train_ids)
idx_val = torch.LongTensor(dev_ids)
idx_test = torch.LongTensor(test_ids)
return adj, features_index, labels, idx_train, idx_val, idx_test
def normalize(mx):
"""Row-normalize sparse matrix"""
rowsum = np.array(mx.sum(1))
r_inv = np.power(rowsum, -1).flatten()
r_inv[np.isinf(r_inv)] = 0.
r_mat_inv = sp.diags(r_inv)
mx = r_mat_inv.dot(mx)
return mx
def accuracy(output, labels):
preds = output.max(1)[1].type_as(labels)
correct = preds.eq(labels).double()
correct = correct.sum()
return correct / len(labels)
def sparse_mx_to_torch_sparse_tensor(sparse_mx):
"""Convert a scipy sparse matrix to a torch sparse tensor."""
sparse_mx = sparse_mx.tocoo().astype(np.float32)
indices = torch.from_numpy(
np.vstack((sparse_mx.row, sparse_mx.col)).astype(np.int64))
values = torch.from_numpy(sparse_mx.data)
shape = torch.Size(sparse_mx.shape)
return torch.sparse.FloatTensor(indices, values, shape)
def load_user_tweet_graph(dataset, elapsed_time, tweets_count):
print('Loading the user_tweet graph of {} dataset ...'.format(dataset))
# if elapsed_time == 3000 and tweets_count == 500:
# path_content = 'dataset/{}/{}_graph.txt'.format(dataset,dataset)
# elif elapsed_time != 3000 and tweets_count == 500:
# path_content = 'dataset/{}/{}_graph_et{}.txt'.format(dataset, dataset, elapsed_time)
# elif elapsed_time == 3000 and tweets_count != 500:
# path_content = 'dataset/{}/{}_graph_tc{}.txt'.format(dataset, dataset, tweets_count)
#
# X_tids = []
# X_uids = []
# with open('dataset/{}/{}.idx.txt'.format(dataset, dataset), 'r') as f:
# line = f.readline()
# X_tids = line.split()
#
# with open(path_content, 'r', encoding='utf-8') as input:
# relation = []
# for line in input.readlines():
# tmp = line.strip().split()
# src = tmp[0]
# X_uids.append(src)
#
# for dst_ids_ws in tmp[1:]:
# dst, w = dst_ids_ws.split(":")
# X_uids.append(dst)
# relation.append([src, dst, w])
#
# X_id = list(set(X_tids + X_uids))
# num_node = len(X_id)
# print(num_node)
# X_id_dic = {id:i for i, id in enumerate(X_id)}
#
# relation = np.array([[X_id_dic[tup[0]], X_id_dic[tup[1]], tup[2]] for tup in relation])
# relation = build_symmetric_adjacency_matrix(relation, shape=(num_node, num_node))
# train_idx = []
# dev_idx = []
# test_idx = []
# with open('dataset/{}/{}.train'.format(dataset, dataset), 'r') as f:
# lines = f.readlines()
# for line in lines:
# idx = line.strip().split()[0]
# train_idx.append(X_id_dic[idx])
# with open('dataset/{}/{}.dev'.format(dataset, dataset), 'r') as f:
# lines = f.readlines()
# for line in lines:
# idx = line.strip().split()[0]
# dev_idx.append(X_id_dic[idx])
# with open('dataset/{}/{}.test'.format(dataset, dataset), 'r') as f:
# lines = f.readlines()
# for line in lines:
# idx = line.strip().split()[0]
# test_idx.append(X_id_dic[idx])
names = ['train', 'dev', 'test', 'adj']
objects = []
for i in range(len(names)):
if elapsed_time == 3000 and tweets_count == 500:
path = 'dataset/{}/ind.{}.user.tweet.{}'.format(dataset, dataset, names[i])
elif elapsed_time != 3000 and tweets_count == 500:
path = 'dataset/{}/ind.{}.user.tweet.{}.et{}'.format(dataset, dataset, names[i], elapsed_time)
elif elapsed_time == 3000 and tweets_count != 500:
path = 'dataset/{}/ind.{}.user.tweet.{}.tc{}'.format(dataset, dataset, names[i], tweets_count)
with open(path, 'rb') as f:
objects.append(jlb.load(f))
train_idx, dev_idx, test_idx, adj = tuple(objects)
adj = adj + adj.T.multiply(adj.T > adj) - adj.multiply(adj.T > adj)
adj = normalize_adj(adj + sp.eye(adj.shape[0]))
relation = sparse_mx_to_torch_sparse_tensor(adj)
train_idx = torch.LongTensor(train_idx)
dev_idx = torch.LongTensor(dev_idx)
test_idx = torch.LongTensor(test_idx)
return train_idx, dev_idx, test_idx, relation
def normalize_adj(mx):
"""Row-normalize sparse matrix"""
rowsum = np.array(mx.sum(1))
r_inv_sqrt = np.power(rowsum, -0.5).flatten()
r_inv_sqrt[np.isinf(r_inv_sqrt)] = 0.
r_mat_inv_sqrt = sp.diags(r_inv_sqrt)
return mx.dot(r_mat_inv_sqrt).transpose().dot(r_mat_inv_sqrt)
def build_symmetric_adjacency_matrix(edges, shape):
adj = sp.coo_matrix((edges[:, 2], (edges[:, 0], edges[:, 1])), shape=shape, dtype=np.float32)
# build symmetric adjacency matrix
adj = adj + adj.T.multiply(adj.T > adj) - adj.multiply(adj.T > adj)
adj = normalize_adj(adj + sp.eye(adj.shape[0]))
# adj = sparse_mx_to_torch_sparse_tensor(adj)
return adj.tocoo()
def evaluation_4class(prediction, y): # 4 dim
prediction = prediction.detach().cpu().numpy()
y = y.detach().cpu().numpy()
TP1, FP1, FN1, TN1 = 0, 0, 0, 0
TP2, FP2, FN2, TN2 = 0, 0, 0, 0
TP3, FP3, FN3, TN3 = 0, 0, 0, 0
TP4, FP4, FN4, TN4 = 0, 0, 0, 0
e, RMSE, RMSE1, RMSE2, RMSE3, RMSE4 = 0.000001, 0.0, 0.0, 0.0, 0.0, 0.0
for i in range(len(y)):
y_i, p_i = list(y[i]), list(prediction[i])
##RMSE
for j in range(len(y_i)):
RMSE += (y_i[j]-p_i[j])**2
RMSE1 += (y_i[0]-p_i[0])**2
RMSE2 += (y_i[1]-p_i[1])**2
RMSE3 += (y_i[2]-p_i[2])**2
RMSE4 += (y_i[3]-p_i[3])**2
## Pre, Recall, F
Act = str(y_i.index(max(y_i))+1)
Pre = str(p_i.index(max(p_i))+1)
#print y_i, p_i
#print Act, Pre
## for class 1
if Act == '1' and Pre == '1': TP1 += 1
if Act == '1' and Pre != '1': FN1 += 1
if Act != '1' and Pre == '1': FP1 += 1
if Act != '1' and Pre != '1': TN1 += 1
## for class 2
if Act == '2' and Pre == '2': TP2 += 1
if Act == '2' and Pre != '2': FN2 += 1
if Act != '2' and Pre == '2': FP2 += 1
if Act != '2' and Pre != '2': TN2 += 1
## for class 3
if Act == '3' and Pre == '3': TP3 += 1
if Act == '3' and Pre != '3': FN3 += 1
if Act != '3' and Pre == '3': FP3 += 1
if Act != '3' and Pre != '3': TN3 += 1
## for class 4
if Act == '4' and Pre == '4': TP4 += 1
if Act == '4' and Pre != '4': FN4 += 1
if Act != '4' and Pre == '4': FP4 += 1
if Act != '4' and Pre != '4': TN4 += 1
## print result
Acc_all = round( float(TP1+TP2+TP3+TP4)/float(len(y)+e), 4 )
Acc1 = round( float(TP1+TN1)/float(TP1+TN1+FN1+FP1+e), 4 )
Prec1 = round( float(TP1)/float(TP1+FP1+e), 4 )
Recll1 = round( float(TP1)/float(TP1+FN1+e), 4 )
F1 = round( 2*Prec1*Recll1/(Prec1+Recll1+e), 4 )
Acc2 = round( float(TP2+TN2)/float(TP2+TN2+FN2+FP2+e), 4 )
Prec2 = round( float(TP2)/float(TP2+FP2+e), 4 )
Recll2 = round( float(TP2)/float(TP2+FN2+e), 4 )
F2 = round( 2*Prec2*Recll2/(Prec2+Recll2+e), 4 )
Acc3 = round( float(TP3+TN3)/float(TP3+TN3+FN3+FP3+e), 4 )
Prec3 = round( float(TP3)/float(TP3+FP3+e), 4 )
Recll3 = round( float(TP3)/float(TP3+FN3+e), 4 )
F3 = round( 2*Prec3*Recll3/(Prec3+Recll3+e), 4 )
Acc4 = round( float(TP4+TN4)/float(TP4+TN4+FN4+FP4+e), 4 )
Prec4 = round( float(TP4)/float(TP4+FP4+e), 4 )
Recll4 = round( float(TP4)/float(TP4+FN4+e), 4 )
F4 = round( 2*Prec4*Recll4/(Prec4+Recll4+e), 4 )
microF = round( (F1+F2+F3+F4)/4,5 )
RMSE_all = round( ( RMSE/len(y) )**0.5, 4)
RMSE_all_1 = round( ( RMSE1/len(y) )**0.5, 4)
RMSE_all_2 = round( ( RMSE2/len(y) )**0.5, 4)
RMSE_all_3 = round( ( RMSE3/len(y) )**0.5, 4)
RMSE_all_4 = round( ( RMSE4/len(y) )**0.5, 4)
RMSE_all_avg = round( ( RMSE_all_1+RMSE_all_2+RMSE_all_3+RMSE_all_4 )/4, 4)
# return ['acc:', Acc_all, 'Favg:',microF, RMSE_all, RMSE_all_avg,
# 'C1:',Acc1, Prec1, Recll1, F1,
# 'C2:',Acc2, Prec2, Recll2, F2,
# 'C3:',Acc3, Prec3, Recll3, F3,
# 'C4:',Acc4, Prec4, Recll4, F4]
return Acc_all, microF, RMSE_all, RMSE_all_avg, Acc1, Prec1, Recll1, F1, Acc2, Prec2, Recll2, F2, Acc3, Prec3,Recll3, F3, Acc4, Prec4, Recll4, F4
def convert_to_one_hot(y, C):
# return np.eye(C)[y.reshape(-1)]
return torch.zeros(y.shape[0], C).scatter_(1, y, 1)