-
Notifications
You must be signed in to change notification settings - Fork 0
/
rec_preprocess.py
207 lines (188 loc) · 9.31 KB
/
rec_preprocess.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
import ujson as json
import os
import pandas as pd
from tqdm import tqdm
from rec_util import dump_pkl
class BasicParser(object):
def __init__(self, file_path, user_in, item_in, T, P, data_type):
print("Generating {} samples...".format(data_type))
user_path = os.path.join(file_path, user_in)
self.ulines = open(user_path, 'r').readlines()
item_path = os.path.join(file_path, item_in)
self.ilines = open(item_path, 'r').readlines()
self.NU = len(self.ulines)
self.NI = len(self.ilines)
self.T = T
self.P = P
print("Initial users {} items {}".format(self.NU, self.NI))
def parse_record(self, user_out, item_out, ulen_out, ilen_out):
urecords, urec_length = [], []
for line in self.ulines:
urecords.append(json.loads(line))
for uid, record in tqdm(enumerate(urecords), total=len(urecords)):
uid = str(uid + 1)
u_len = len(record[uid])
if u_len == 0:
print(uid)
if u_len >= self.T:
record[uid] = record[uid][:self.T]
u_len = self.T
else:
for i in range(self.T - u_len):
record[uid].append([0])
urec_length.append(u_len)
irecords, irec_length = [], []
for line in self.ilines:
irecords.append(json.loads(line))
for iid, record in tqdm(enumerate(irecords), total=len(irecords)):
iid = str(iid + 1)
i_len = len(record[iid])
if i_len == 0:
print(iid)
if i_len >= self.T:
record[iid] = record[iid][:self.T]
i_len = self.T
else:
for i in range(self.T - i_len):
record[iid].append([0])
irec_length.append(i_len)
dump_pkl(user_out, urecords)
dump_pkl(item_out, irecords)
dump_pkl(ulen_out, urec_length)
dump_pkl(ilen_out, irec_length)
def parse_set(self, file_path, name_in, name_out, data_type, task):
print("Generating {} samples...".format(data_type))
full_path = os.path.join(file_path, name_in)
raw = pd.read_csv(full_path, sep=',')
uids, iids, labels = [], [], []
for i, row in raw.iterrows():
uid, iid, label = row['uids'], row['iids'], row['labels']
if task == 'static':
uids.append([uid] * self.T)
iids.append([iid] * self.T)
elif task == 'dynamic':
uids.append([t * self.NU + uid for t in range(self.T)])
iids.append([t * self.NI + iid for t in range(self.T)])
else:
uids.append([t % self.P * self.NU + uid for t in range(self.T)])
iids.append([t % self.P * self.NI + iid for t in range(self.T)])
labels.append(label)
processed = {'uids': uids, 'iids': iids, 'labels': labels}
dump_pkl(name_out, processed)
class DynamicParser(BasicParser):
def parse_record(self, user_out, item_out, ulen_out, ilen_out):
urecords, urec_length = [], []
for line in self.ulines:
urecords.append(json.loads(line))
for uid, record in tqdm(enumerate(urecords), total=len(urecords)):
uid = str(uid + 1)
u_len = len(record[uid])
# u_raw_length.append(u_len)
if u_len >= self.T:
record[uid] = record[uid][:self.T]
u_len = self.T
else:
for i in range(self.T - u_len):
record[uid].append([0])
for t in range(u_len):
# record[uid][t] 第t月与user有交互的item list
for idx, ut_i in enumerate(record[uid][t]):
record[uid][t][idx] = t * self.NI + ut_i # 按第t月寻找组 按id偏移 方便id对应到item的第t个月的embedding
urec_length.append(u_len)
irecords, irec_length = [], []
for line in self.ilines:
irecords.append(json.loads(line))
for iid, record in tqdm(enumerate(irecords), total=len(irecords)):
iid = str(iid + 1)
i_len = len(record[iid])
# i_raw_length.append(i_len)
if i_len >= self.T:
record[iid] = record[iid][: self.T]
i_len = self.T
else:
for i in range(self.T - i_len):
record[iid].append([0])
for t in range(i_len):
for idx, it_u in enumerate(record[iid][t]):
record[iid][t][idx] = t * self.NU + it_u
irec_length.append(i_len)
dump_pkl(user_out, urecords)
dump_pkl(item_out, irecords)
dump_pkl(ulen_out, urec_length)
dump_pkl(ilen_out, irec_length)
class PeriodParser(BasicParser):
def parse_record(self, user_out, item_out, ulen_out, ilen_out):
urecords, urec_length = [], []
for line in self.ulines: # each line is the records of an user, key=user_id value=records
urecords.append(json.loads(line))
for uid, record in tqdm(enumerate(urecords), total=len(urecords)):
uid = str(uid + 1)
u_len = len(record[uid]) # each sublist is the records in a month
# u_raw_length.append(u_len)
if u_len >= self.T:
record[uid] = record[uid][:self.T]
u_len = self.T
else:
for i in range(self.T - u_len):
record[uid].append([0])
for t in range(u_len):
for idx, ut_i in enumerate(record[uid][t]):
record[uid][t][idx] = t % self.P * self.NI + ut_i # 按第t月寻找组 按id偏移
urec_length.append(u_len)
irecords, irec_length = [], []
for line in self.ilines:
irecords.append(json.loads(line))
for iid, record in tqdm(enumerate(irecords), total=len(irecords)):
iid = str(iid + 1)
i_len = len(record[iid])
# i_raw_length.append(i_len)
if i_len >= self.T:
record[iid] = record[iid][:self.T]
i_len = self.T
else:
for i in range(self.T - i_len):
record[iid].append([0])
for t in range(i_len):
for idx, it_u in enumerate(record[iid][t]):
record[iid][t][idx] = t % self.P * self.NU + it_u
irec_length.append(i_len)
dump_pkl(user_out, urecords)
dump_pkl(item_out, irecords)
dump_pkl(ulen_out, urec_length)
dump_pkl(ilen_out, irec_length)
def run_prepare(config, flags):
if config.dynamic:
parser = DynamicParser(config.raw_dir, config.user_record_file, config.item_record_file,
config.T, config.P, 'record')
parser.parse_record(flags.user_record_file, flags.item_record_file,
flags.user_length_file, flags.item_length_file)
task = 'dynamic'
# num_user, num_item = dynamic_parser.NU, dynamic_parser.NI
# parse_dynamic_set(config.raw_dir, config.train_file, config.T, num_user, num_item, flags.train_file, 'train')
# parse_dynamic_set(config.raw_dir, config.valid_file, config.T, num_user, num_item, flags.valid_file, 'valid')
# parse_dynamic_set(config.raw_dir, config.test_file, config.T, num_user, num_item, flags.test_file, 'test')
elif config.period:
parser = PeriodParser(config.raw_dir, config.user_record_file, config.item_record_file,
config.T, config.P, 'record')
parser.parse_record(flags.user_record_file, flags.item_record_file,
flags.user_length_file, flags.item_length_file)
task = 'period'
# num_user, num_item = period_parser.NU, period_parser.NI
# parse_period_set(config.raw_dir, config.train_file, config.T, config.P, num_user, num_item, flags.train_file,
# 'train')
# parse_period_set(config.raw_dir, config.valid_file, config.T, config.P, num_user, num_item, flags.valid_file,
# 'valid')
# parse_period_set(config.raw_dir, config.test_file, config.T, config.P, num_user, num_item, flags.test_file,
# 'test')
else:
parser = BasicParser(config.raw_dir, config.user_record_file, config.item_record_file,
config.T, config.P, 'record')
parser.parse_record(flags.user_record_file, flags.item_record_file,
flags.user_length_file, flags.item_length_file)
task = 'static'
# parse_set(config.raw_dir, config.train_file, config.T, flags.train_file, 'train')
# parse_set(config.raw_dir, config.valid_file, config.T, flags.valid_file, 'valid')
# parse_set(config.raw_dir, config.test_file, config.T, flags.test_file, 'test')
parser.parse_set(config.raw_dir, config.train_file, flags.train_file, 'train', task)
parser.parse_set(config.raw_dir, config.valid_file, flags.valid_file, 'valid', task)
parser.parse_set(config.raw_dir, config.test_file, flags.test_file, 'test', task)