-
Notifications
You must be signed in to change notification settings - Fork 1
/
data_utils.py
197 lines (157 loc) · 6.99 KB
/
data_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
import ast
import csv
import itertools
import numpy as np
import scipy.sparse as sp
from sklearn.cluster import KMeans
from sklearn.decomposition import TruncatedSVD
from ki_means import KIMeans
def parse(data):
for line in data:
if not line:
continue
user_id = line[0]
for anime in line[1:]:
anime_norm_id, anime_score = ast.literal_eval(anime)
yield int(user_id) - 1, int(anime_norm_id) - 1, int(anime_score)
def get_dimensions(train_data, test_data):
print("Getting dimensions")
user_ids = set()
anime_ids = set()
for uid, aid, _ in itertools.chain(train_data, test_data):
user_ids.add(uid)
anime_ids.add(aid)
rows = max(user_ids) + 1
cols = max(anime_ids) + 1
return rows, cols
def build_interaction_matrix(rows, cols, data, min_rating):
print("Building interaction matrix")
matrix = sp.lil_matrix((rows, cols), dtype=np.int32)
for user_id, anime_id, score in data:
if score >= min_rating:
matrix[user_id, anime_id] = score
return matrix.tocoo()
def build_item_metadata_matrix(num_items, metadata):
print("Building metadata matrix")
header = next(metadata)
n_items, n_features = [int(i) for i in header]
assert num_items == n_items
id_features = sp.identity(num_items, format='csr', dtype=np.float32)
features = sp.lil_matrix((num_items, n_features), dtype=np.float32)
for line in metadata:
# blank line
if not line:
continue
# no feature
if len(line) < 2:
continue
anime_id = int(line[0]) - 1
for feature in line[1:]:
features[anime_id, int(feature) - 1] = 1.0
return id_features, features.tocsr()
def get_user_clustering_dimensions(data):
print("Getting clustering dimensions")
user_ids = set()
clusters = set()
for uid, _, c in data:
user_ids.add(int(uid))
clusters.add(int(c))
rows = max(user_ids)
cols = max(clusters) + 1
return rows, cols
def build_user_feature_matrix(num_users, user_data, clusters):
print("Building user feature matrix")
id_features = sp.identity(num_users, format='csr', dtype=np.float32)
features = sp.lil_matrix((num_users, clusters), dtype=np.float32)
for line in user_data:
# blank line
if not line:
continue
# no feature
if len(line) < 3:
continue
user_id = int(line[0]) - 1
features[user_id, int(line[2])] = 1.0
return id_features, features.tocsr()
def fetch_anime(train_file, test_file, user_features_file=None, features_file=None, min_score=0):
print("Start Fetching")
user_feature_matrix = None
item_feature_matrix = None
with open(train_file, 'r', encoding='utf-8') as train_f:
with open(test_file, 'r', encoding='utf-8') as tests_f:
train = csv.reader(train_f, delimiter=',')
test = csv.reader(tests_f, delimiter=',')
num_users, num_items = get_dimensions(parse(train), parse(test))
train_f.seek(0)
tests_f.seek(0)
train_matrix = build_interaction_matrix(num_users, num_items, parse(train), min_score)
test_matrix = build_interaction_matrix(num_users, num_items, parse(test), min_score)
assert train_matrix.shape == test_matrix.shape
if user_features_file:
with open(user_features_file, 'r', encoding='utf-8') as user_features_f:
features = csv.reader(user_features_f, delimiter=',')
n_users, clusters = get_user_clustering_dimensions(features)
assert num_users == n_users
user_features_f.seek(0)
id_features, feature_matrix = build_user_feature_matrix(num_users, features, clusters)
user_feature_matrix = sp.hstack([id_features, feature_matrix]).tocsr()
if features_file:
with open(features_file, 'r', encoding='utf-8') as item_features_f:
features = csv.reader(item_features_f, delimiter=',')
id_features, feature_matrix = build_item_metadata_matrix(num_items, features)
item_feature_matrix = sp.hstack([id_features, feature_matrix]).tocsr()
return {'train_set': train_matrix,
'test_set': test_matrix,
'user_features': user_feature_matrix,
'item_features': item_feature_matrix}
def fetch_all_data(data_file, min_score):
print("Reading interaction file")
with open(data_file, 'r', encoding='utf-8') as f:
data = csv.reader(f, delimiter=',')
num_users, num_items = get_dimensions(parse(data), [])
f.seek(0)
print("Building matrix...")
interaction_matrix = build_interaction_matrix(num_users, num_items, parse(data), min_score)
return interaction_matrix
def run_kmeans(data, k, user_ids_file, out_file, debug=False):
print("Starting K means")
if debug:
labeler = KMeans(n_clusters=k, random_state=0)
else:
labeler = KMeans(n_clusters=k)
print("Fitting...")
labeler.fit(data.tocsr())
with open(user_ids_file, 'r', encoding='utf-8') as f:
with open(out_file, 'w', encoding='utf-8', newline="\n") as out:
reader = csv.reader(f)
writer = csv.writer(out)
for n_line, line in enumerate(reader):
writer.writerow(line + [labeler.labels_[n_line]])
def run_lsa_kmeans(data, k, user_ids_file, out_file, reduce_dim_to=1000, iter=10, debug=False):
print("Starting LSA")
if debug:
svd = TruncatedSVD(n_components=reduce_dim_to, n_iter=iter, random_state=0)
else:
svd = TruncatedSVD(n_components=reduce_dim_to, n_iter=iter)
print("Fitting LSA...")
reduced_data = svd.fit_transform(data)
run_kmeans(sp.coo_matrix(reduced_data), k, user_ids_file, out_file, debug)
def run_kimeans(data, k, iter, user_ids_file, out_file, debug=False):
print("Starting Ki means")
ki = KIMeans(k, iter)
print("Fitting...")
ki.fit(np.array(data.todense()))
clusters = ki.get_correlations()[1]
with open(user_ids_file, 'r', encoding='utf-8') as f:
with open(out_file, 'w', encoding='utf-8', newline="\n") as out:
reader = csv.reader(f)
writer = csv.writer(out)
for n_line, line in enumerate(reader):
writer.writerow(line + [int(clusters[n_line][0])])
if __name__ == "__main__":
dicta = fetch_anime(train_file='user_interactions_train.csv', test_file='user_interactions_test.csv',
user_features_file='kmeans_score0_out.csv', features_file='anime_director_ind.csv', min_score=0)
print(dicta)
# anime_data = fetch_all_data('user_interactions.csv', 0)
# run_lsa_kmeans(anime_data, 50, user_ids_file='Data/users_id.csv', out_file='kmeans_lsa_dim100_out.csv', debug=True, reduce_dim_to=100)
# run_kimeans(anime_data, 50, iter=100, user_ids_file='Data/users_id.csv', out_file='kimeans_out.csv', debug=True)