forked from Jhy1993/HAN
-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess_dblp.py
92 lines (79 loc) · 2.56 KB
/
preprocess_dblp.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
all_keyword_list = []
with open(p_path, 'r', encoding='utf-8') as f:
for line in f.readlines():
line = line.lower()
line = line.strip('\n').split('\t')
# print(line)
line[1] = re.findall('[a-zA-Z0-9]+', line[1])
all_keyword_list.extend(
[tmp_key for tmp_key in line[1] if tmp_key not in sp_word])
cnt = collections.Counter(all_keyword_list)
selected_keyword = []
for k, v in cnt.items():
if v > 50:
selected_keyword.append(k)
# 最终选择 334 个selected_keyword
with open(p_path, 'r', encoding='utf-8') as f:
for line in f.readlines():
line = line.lower()
line = line.strip('\n').split('\t')
# print(line)
line[1] = re.findall('[a-zA-Z0-9]+', line[1])
paper2key[str('P' + line[0])] = [tmp_key for tmp_key in line[1] if
tmp_key in selected_keyword]
#
author_word = []
from sklearn.preprocessing import OneHotEncoder, LabelBinarizer, MultiLabelBinarizer
for author in selected_author:
tmp = []
for pi in adj_dict_ap[author]:
tmp.extend(paper2key[pi])
author_word.append(tmp)
ohe = MultiLabelBinarizer()
author2feature = ohe.fit_transform(author_word)
idx2term = {}
with open(t_path, 'r') as f:
for line in f.readlines():
line = line.strip('\n').split('\t')
idx2term[line[0]] = line[1]
term2idx = {v: k for k, v in idx2term.items()}
idx2sp = {}
for k, v in idx2term.items():
if v in stopwords.words('english'):
idx2sp[k] = v
def split_idx(author_label, train_size, val_size):
train_per_cls = int(train_size / 4)
val_per_cls = int(val_size / 4)
y = np.argmax(author_label, axis=1)
train, val, test = [], [], []
k0, k1, k2, k3 = 0, 0, 0, 0
for i in range(y.shape[0]):
if y[i] == 0 and k0 < train_per_cls:
train.append(i)
k0 += 1
elif y[i] == 0 and train_per_cls <= k0 < train_per_cls + val_per_cls:
val.append(i)
k0 += 1
elif y[i] == 1 and k1 < train_per_cls:
train.append(i)
k1 += 1
elif y[i] == 1 and train_per_cls <= k1 < train_per_cls + val_per_cls:
val.append(i)
k1 += 1
elif y[i] == 2 and k2 < train_per_cls:
train.append(i)
k2 += 1
elif y[i] == 2 and train_per_cls <= k2 < train_per_cls + val_per_cls:
val.append(i)
k2 += 1
elif y[i] == 3 and k3 < train_per_cls:
train.append(i)
k3 += 1
elif y[i] == 3 and train_per_cls <= k3 < train_per_cls + val_per_cls:
val.append(i)
k3 += 1
else:
test.append(i)
print('train_size: {}, val_szie: {}, test_size: {}'.format(
len(train), len(val), len(test)))
return train, val, test