-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataloader.py
257 lines (226 loc) · 9.99 KB
/
dataloader.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
# -*- coding: utf-8 -*-
"""
Created on Sat Jan 4 18:27:48 2020
@author: Eugene
"""
import pandas as pd
import numpy as np
from collections import defaultdict
import pickle
from tqdm import tqdm
tqdm.pandas()
def gen_edges(target_rank, save=False, pp_only=False):
p_p = pd.read_pickle('paper_paper.pkl')
p_v = pd.read_pickle('paper_venue.pkl')
p_a = pd.read_pickle('paper_author.pkl')
keyword_list = np.load('keyword_list.npy')
data = pd.read_pickle('dblp_0105.pkl')
paper_time_step = data['time_step']
# target_rank = 162 # default 281
target_year = 2018
hot_rate = 5
good_author_num = 2
last_pv_id = np.max(p_v['new_venue_id'])
last_pv_id += 1
# venue year
data2 = data.copy()
venue_year = pd.DataFrame(data2[['venue_year','time_step']].drop_duplicates().reset_index(drop=True))
venue_year = venue_year.sort_values(by=['time_step'])
venue_year['new_venue_year_id'] = np.arange(last_pv_id,last_pv_id+len(venue_year))
paper_venue_year = pd.merge(data,venue_year)[['new_papr_id','new_venue_year_id','time_step']]
paper_venue_year = paper_venue_year.reset_index(drop=True)
# origin_venue = p_v['new_venue_id'].copy().drop_duplicates()
'''paper發在哪一年的哪個venue裡面，2018KDD和2019KDD是兩個不同的venue'''
p_vy = paper_venue_year.copy()
pvy = p_vy[['new_papr_id','new_venue_year_id']]
paper_year = p_v[['new_papr_id','time_step','year']]
pvy_new = pd.merge(paper_year,pvy)
pvy_group = pvy_new.groupby('year')
c = 8 # year_venue relation
pvy_new = pd.DataFrame()
for i in range(2011,2020):
group_temp = pvy_group.get_group(i)
group_temp = group_temp.reset_index(drop=True)
rel = pd.DataFrame(np.ones(len(group_temp))*c)
rel.columns = ['rel']
group_temp = pd.concat((group_temp,rel),axis=1)
group_temp = group_temp[['new_papr_id','new_venue_year_id','rel','time_step']]
group_temp.columns = ['head','tail','rel','time_step']
c += 1
pvy_new = pvy_new.append(group_temp)
pvy = pvy_new[pvy_new['time_step']<target_rank][['head', 'tail', 'rel']]
print(pvy['tail'].unique().shape)
#venue year venue
'''EX:2018KDD和2019KDD都屬於KDD這個venue'''
venue_year_venue = pd.merge(paper_venue_year,p_v)[['new_venue_year_id','new_venue_id','time_step']].drop_duplicates()
venue_year_venue = venue_year_venue.reset_index(drop=True)
last_vy_id = np.max(venue_year_venue['new_venue_year_id'])
last_vy_id += 1
vy_v = venue_year_venue.copy()
vyv = vy_v[vy_v['time_step']<target_rank][['new_venue_year_id','new_venue_id']]
vyv = vyv.reset_index(drop=True)
rel = pd.DataFrame(np.ones(len(vyv))*7) # venue id is 7
vyv = pd.concat((vyv,rel),axis=1)
vyv.columns = ['head','tail','rel']
# keyword
'''paper裡面有哪些keyword的關係'''
keyword_list = list(dict.fromkeys(keyword_list))
keydict = {keyword_list[i]: i+last_vy_id for i in range(0, len(keyword_list))}
keyword = data['keyword']
def get_keyword_id(df):
key_id = []
for i in df:
key_id.append(keydict[i])
return key_id
paper_keyword = []
for i in range(len(keyword)):
for k in keyword.iloc[i]:
paper_keyword.append([i,keydict[k],paper_time_step.iloc[i]])
paper_keyword = pd.DataFrame(paper_keyword)
paper_keyword.columns = ['new_papr_id','keyword_id','time_step']
pk = paper_keyword[paper_keyword['time_step']<target_rank][['new_papr_id','keyword_id']]
pk = pk.reset_index(drop=True)
rel = pd.DataFrame(np.ones(len(pk))*6)
pk = pd.concat((pk,rel),axis=1)
pk.columns = ['head','tail','rel']
#hot paper
#target_REF = p_p[p_p['time_step']<target_rank]
#cited_value = target_REF['new_cited_papr_id'].value_counts()
#cited_value = pd.DataFrame(cited_value)
#cited_value['pub_year'] = target_year - data.iloc[cited_value.index]['year'] + 1
#cited_value['cited_rate'] = cited_value['new_cited_papr_id']/cited_value['pub_year']
#good_paper = list(cited_value[cited_value['cited_rate']>hot_rate].index)
#paper hot 先篩reference時間再找hot paper>0.5
#pp = p_p[p_p['time_step']<target_rank][['new_papr_id','new_cited_papr_id']]
#
#def choose_hot_paper(df):
# if df in good_paper:
# return df
# else:
# return None
#
#print('process hot cite')
#pp_hot = pd.DataFrame(pp['new_cited_papr_id'].progress_apply(choose_hot_paper))
#pp_hot.columns = ['good_paper']
#pp_hot = pd.concat((pp,pp_hot),axis=1).dropna()
#pp_hot = pp_hot.drop(columns=['new_cited_papr_id'])
#pp_hot = pp_hot.reset_index(drop=True)
#pp_hot.columns = ['head','tail']
#ph = pp_hot.copy()
#rel = pd.DataFrame(np.ones(len(ph))*2)
#ph = pd.concat((ph,rel),axis=1)
#ph.columns = ['head','tail','rel']
#delete author 先照時間篩選pa 再找寫>1篇的作者
'''paper的author是誰的關係'''
pa = p_a[p_a['time_step']<target_rank][['new_papr_id','new_author_id']]
author_value = pa['new_author_id'].value_counts()
author_value = pd.DataFrame(author_value)
bad_author = list(author_value[author_value['new_author_id']<=good_author_num].index)
good_author = list(author_value[author_value['new_author_id']>good_author_num].index)
pa_new = pa[~pa['new_author_id'].isin(bad_author)]
pa_new.columns = ['head','tail']
pa_new = pa_new.reset_index(drop=True)
rel = pd.DataFrame(np.ones(len(pa_new))) # author relation is 1
pa = pd.concat((pa_new,rel),axis=1)
pa.columns = ['head','tail','rel']
#self cite 會因為不同時間作者不同而改變
#paper_author = defaultdict(list)
#
#for i in pa_new.values:
# paper_author[i[0]].append(i[1])
#
#exist_paper = list(paper_author.keys())
#
#def find_self_cite(cite,cited):
# if cite in exist_paper and cited in exist_paper:
# if len(list(set(paper_author[cite]).intersection(set(paper_author[cited]))))>0:
# return 1
# return 0
# else:
# return 0
#
#print('process self cite')
#pp['self_cite'] = pp.progress_apply(lambda row: find_self_cite(row['new_papr_id'], row['new_cited_papr_id']), axis=1)
#pp_self_cite = pp[pp['self_cite']==1]
#pp_self_cite = pp_self_cite.drop(columns=['self_cite'])
#pp_self_cite.columns = ['head','tail']
#pp_self_cite = pp_self_cite.reset_index(drop=True)
#pself = pp_self_cite.copy()
#rel = pd.DataFrame(np.ones(len(pself))*4)
#pself = pd.concat((pself,rel),axis=1)
#pself.columns = ['head','tail','rel']
#newest cite
#pp_newest = p_p[['new_papr_id','new_cited_papr_id','year','time_step']].reset_index(drop=True)
#pp_newest.columns = ['head','tail','year','time_step']
#cite_year = data.iloc[pp_newest['tail'].values]['year'].reset_index(drop=True)
#pp_newest['cite_year'] = cite_year
#pp_newest['cite_newest'] = pp_newest['year'] - pp_newest['cite_year']
#pp_newest_cite = pp_newest[(pp_newest['cite_newest']<=1) & (pp_newest['cite_newest']>=0)][['head','tail','time_step']]
#pp_new = pp_newest_cite[pp_newest_cite['time_step']<target_rank].drop(columns=['time_step'])
#pp_new = pp_new.reset_index(drop=True)
#
#pnew = pp_new.copy()
#rel = pd.DataFrame(np.ones(len(pnew))*3)
#pnew = pd.concat((pnew,rel),axis=1)
#pnew.columns = ['head','tail','rel']
#survey cite pp已篩過時間
#title = pd.DataFrame(data['title'].str.lower())
#title['if suevey or not'] = title['title'].apply(lambda x : 1 if 'survey' in x else 0)
#title['if review or not'] = title['title'].apply(lambda x : 1 if 'review on' in x or 'review of' in x else 0)
#survey_paper_list = pd.concat((data['new_papr_id'],title),axis=1)
#survey_paper_list = survey_paper_list[(survey_paper_list['if review or not']==1) | (survey_paper_list['if suevey or not'] ==1)]['new_papr_id'].tolist()
#
#def choose_survey_paper(df):
# if df in survey_paper_list:
# return df
# else:
# return None
#
#print('process survey cite')
#pp_survey = pd.DataFrame(pp['new_cited_papr_id'].progress_apply(choose_survey_paper))
#pp_survey.columns = ['survey_paper']
#pp_survey = pd.concat((pp,pp_survey),axis=1).dropna()
#pp_survey = pp_survey.drop(columns=['new_cited_papr_id','self_cite'])
#pp_survey = pp_survey.reset_index(drop=True)
#psurvey = pp_survey.copy()
#rel = pd.DataFrame(np.ones(len(psurvey))*5)
#psurvey = pd.concat((psurvey,rel),axis=1)
#psurvey.columns = ['head','tail','rel']
#pp
'''paper cite paper的關係'''
if pp_only:
pp = p_p[p_p['time_step'] == target_rank][['new_papr_id', 'new_cited_papr_id']]
else:
pp = p_p[p_p['time_step'] < target_rank-1][['new_papr_id', 'new_cited_papr_id']]
pp = pp.reset_index(drop=True)
rel = pd.DataFrame(np.zeros(len(pp)))
pp = pd.concat((pp,rel),axis=1)
pp.columns = ['head', 'tail', 'rel']
if pp_only:
return pp
# edge list
# all_edge = pd.concat([pp,pa,ph,pnew,pself,psurvey,pk,vyv,pvy])
all_edge = pd.concat([pp,pa,pk,vyv,pvy])
all_edge = all_edge.drop_duplicates()
all_edge = all_edge.reset_index(drop=True)
if save:
with open('all_edge_'+str(target_rank)+'.pkl', 'wb') as file:
pickle.dump(all_edge, file)
# map to new id
# index = np.arange(len(list(set(pd.unique(all_edge['tail'])).union(set(pd.unique(all_edge['head']))))))
# content = np.array(list(set(pd.unique(all_edge['tail'])).union(set(pd.unique(all_edge['head'])))))
# new_index_dict = dict(zip(content, index))
# all_edge['head'] = all_edge['head'].map(new_index_dict)
# all_edge['tail'] = all_edge['tail'].map(new_index_dict)
return all_edge
# def compute_node(el):
# h = list(el['head'].values)
# t = list(el['tail'].values)
# h.extend(t)
# h = list(set(h))
# print('node nums:',len(h))
# print('max node id:',int(max(h)))
# print('edge nums:',len(el))
# return int(max(h)),len(h)
#
# max_id,num_nodes = compute_node(all_edge)