-
Notifications
You must be signed in to change notification settings - Fork 1
/
utils.py
208 lines (174 loc) · 9.13 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
import os
import re
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
import pickle
import networkx as nx
import gzip
import json
import os
import pandas as pd
pd.options.display.float_format = '{:,}'.format
# ------------BOOKS--------------------
def noise_sparsity(C, noise, sparsity):
C[np.random.rand(*C.shape) > sparsity] = 0
noisy_pair = np.random.rand(*C.shape) < noise
C[(C == -1) & noisy_pair] = 1
C[(C == 1) & noisy_pair] = -1
return C
def df_to_numpy(path1="./Books/subdataset/books_2D.csv",
path2="./Books/subdataset/books_info.csv",
outpath="./Books/subdataset/C.pickle",
noise=0, sparsity=1):
np.random.seed(123456)
small_df = pd.read_csv(path1)
genre_df = pd.read_csv(path2)
userId = np.array(small_df["userId"])
movieId = np.array(small_df["movieId"])
rating = np.array(small_df["rating"])
label_genre = np.array(genre_df["genres"])
movieId_reduce = np.array(genre_df["movieId"])
# movieId_reduce define the order. label_genre and C[:, j] are order in the same way
n_user = len(np.unique(userId))
n_movie = len(movieId_reduce)
C = np.empty((n_user, n_movie, n_movie))
# Label_user define the order. The label returned and the C[i] are order in the same way.
label_user = np.unique(userId)
real_sparsity = 0
for i, label_i in enumerate(label_user):
rating_i = np.zeros(n_movie) + np.percentile(rating[userId == label_i], 33)
for j, label_j in enumerate(movieId_reduce):
if ((userId == label_i) & (movieId == label_j)).sum() >= 1:
rating_i[j] = rating[(userId == label_i) & (movieId == label_j)]
real_sparsity += 1
C[i] = (2 * (rating_i[:, np.newaxis] >= rating_i[np.newaxis, :]) - 1) - \
((rating_i[:, np.newaxis] == rating_i[np.newaxis, :]) * 1)
C[i] = noise_sparsity(C[i], noise=noise, sparsity=sparsity)
real_sparsity /= len(label_user) * len(movieId_reduce)
dict_pickle = {"C": [C],
"True_label": [label_user, label_genre],
"T_pos": [[0, 1, 1]],
"noise": noise,
"sparsity": sparsity,
"real_sparsity": real_sparsity}
with open(outpath, 'wb') as handle:
pickle.dump(dict_pickle, handle)
def assign_label(movies_df, df, genres, n_per_class_):
movies, count = np.unique(df["movieId"], return_counts=True)
movies_with_more_comments = count.argsort()[::-1]
x_return = np.zeros(len(movies), dtype=int)
movies_genre = np.array(movies_df["genres"])
movies_Id = np.array(movies_df["movieId"])
for m in range(len(movies)):
pos_movie_m = np.where(movies_Id == movies[movies_with_more_comments[m]])
assert len(pos_movie_m) == 1
pos_movie_m = pos_movie_m[0]
x = movies_genre[pos_movie_m][0]
movies_genres = x.split("|")
a = []
for i in range(len(genres)):
for j in range(len(genres[i])):
if genres[i][j] in movies_genres:
a.append(i)
break
if len(a) == 1 and n_per_class_[a[0]] > 0:
n_per_class_[a[0]] -= 1
x_return[pos_movie_m] = a[0]
elif len(a) > 1 and n_per_class_[a[1]] > 0 and "War" in genres[1] and "Thriller" in genres[0]:
# Special case for War Western vs thriller
n_per_class_[a[1]] -= 1
x_return[pos_movie_m] = a[1]
else:
x_return[pos_movie_m] = -1
return x_return, n_per_class_
def select_same_labels(df, movies_df,
genres=["Mystery_Thriller_Crime_Drama", "Fantasy_Sci-Fi"],
n_per_class=[10, 10]):
movies_genre = [genre.split("_") for genre in genres]
n_per_class_ = n_per_class.copy()
new_list_genre, n_per_class_ = assign_label(movies_df=movies_df, df=df,
genres=movies_genre, n_per_class_=n_per_class_)
movies_df = movies_df.rename(columns={"genres": "init_genres"})
movies_df.insert(2, "genres", new_list_genre, True)
if sum(n_per_class_) > 0:
print("Warning: Not all the points per classes are selected", n_per_class_)
movies_df = movies_df.loc[movies_df["genres"].isin(range(len(genres)))]
df = df.loc[df["movieId"].isin(movies_df["movieId"])]
return df, movies_df, n_per_class_
def create_movies_dataset(path="./Movies/subdataset/", n_user=100, n_movie=100, n_per_class=[10, 10],
genres=["Mystery_Thriller_Crime_Drama", "Fantasy_Sci-Fi"],
name="", time_split=True, additional_path_save="", rdm_seed=123456):
np.random.seed(rdm_seed)
small_df = pd.read_csv(path + '../ratings' + '.csv')
movies, count = np.unique(small_df["movieId"], return_counts=True)
movies_with_more_comments = count.argsort()[-n_movie:][::-1]
small_df = small_df.loc[small_df['movieId'].isin(movies[movies_with_more_comments])]
movies_df = pd.read_csv(path + '../movies' + '.csv')
movies_df = movies_df.loc[movies_df["movieId"].isin(movies[movies_with_more_comments])]
# Chose the users that comment the most with respect to the movies chosen.
users, count = np.unique(small_df["userId"], return_counts=True)
users_with_more_comments = count.argsort()[-n_user:][::-1]
small_df = small_df.loc[small_df['userId'].isin(users[users_with_more_comments])]
if time_split:
movies_title = np.array(movies_df["title"])
movies_date = []
for i in movies_title:
try:
movies_date.append(int(re.findall("\([0-9]+\)", i)[-1][1:-1]))
except:
movies_date.append(2018)
movies_date = np.array(movies_date)
movies_df["date"] = movies_date
first_half = movies_date >= np.median(movies_date)
first_half = movies_df.loc[first_half, "title"]
movies_df_ = movies_df.loc[movies_df["title"].isin(first_half)]
small_df_ = small_df.loc[small_df["movieId"].isin(movies_df_["movieId"])]
small_df_, movies_df_, n_per_class_ = select_same_labels(small_df_, movies_df_,
genres=genres,
n_per_class=n_per_class)
movies_df_.to_csv(path + additional_path_save + "movies_info" + name + "1.csv")
small_df_.to_csv(path + additional_path_save + "movies_2D" + name + "1.csv")
second_half = np.array([i for i in movies_title if i not in first_half])
# second_half = movies_date < np.median(movies_date)
# small_df["timestamp"] < np.median(small_df["timestamp"])
movies_df_ = movies_df.loc[movies_df["title"].isin(second_half)]
small_df_ = small_df.loc[small_df["movieId"].isin(movies_df_["movieId"])]
# Chose a sub sample of movies by looking at the labels
small_df_, movies_df_, n_per_class_ = select_same_labels(small_df_, movies_df_,
genres=genres,
n_per_class=n_per_class)
movies_df_.to_csv(path + additional_path_save + "movies_info" + name + "2.csv")
small_df_.to_csv(path + additional_path_save + "movies_2D" + name + "2.csv")
else:
# Chose a sub sample of movies by looking at the labels
small_df, movies_df, n_per_class_ = select_same_labels(small_df, movies_df,
genres=genres,
n_per_class=n_per_class)
movies_df.to_csv(path + additional_path_save + "movies_info" + name + ".csv")
small_df.to_csv(path + additional_path_save + "movies_2D" + name + ".csv")
def create_dataset(rdm_seed):
genres = [["Thriller_Crime_Drama", "Fantasy_Sci-Fi"],
["Thriller_Crime_Drama", "Children's_Animation"],
["Thriller_Crime_Drama", "War_Western"],
["Fantasy_Sci-Fi", "Children's_Animation"],
["Fantasy_Sci-Fi", "War_Western"],
["Children's_Animation", "War_Western"]
]
names = ["_" + genre[0][0] + "_" + genre[1][0] for genre in genres]
n_movies = [3000, 3000, 5000, 5000, 5000, 5000]
for i, name in enumerate(names):
create_movies_dataset(path="./Movies/subdataset/", n_user=100, n_movie=n_movies[i], n_per_class=[100, 100],
name=name, genres=genres[i], rdm_seed=rdm_seed)
noise = 0
sparsity = 1
df_to_numpy(path1="./Movies/subdataset/movies_2D" + name + "1.csv",
path2="./Movies/subdataset/movies_info" + name + "1.csv",
outpath="./Movies/subdataset/C" + name + "1.pickle",
noise=noise, sparsity=sparsity)
df_to_numpy(path1="./Movies/subdataset/movies_2D" + name + "2.csv",
path2="./Movies/subdataset/movies_info" + name + "2.csv",
outpath="./Movies/subdataset/C" + name + "2.pickle",
noise=noise, sparsity=sparsity)