-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathDatasetPrep.py
269 lines (207 loc) · 10.8 KB
/
DatasetPrep.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
258
259
260
261
262
263
264
265
266
267
# -*- coding: utf-8 -*-
"""
Created on Wed Nov 21 13:41:59 2018
@author: Mehmet
"""
import numpy as np
#from lightfm.datasets import fetch_stackexchange
import time
from utils import softmax
import matplotlib.pyplot as plt
import os
dirname = os.path.dirname(__file__)
class DatasetPrep(object):
def __init__(self):
self.I = 0
self.J = 0
self.Di = 0
self.Du = 0
self.Mi = 0
self.Mu = 0
self.FeatureUser = []
self.FeatureItem = []
self.ratedata = []
def dataset_stat(self):
R, O = rate_to_matrix(self.ratedata, self.I, self.J)
print("Rating Mean: {:.2f}, Std: {:.2f}, Min: {:.2f}, Max: {:.2f}".format(self.ratedata[:, 2].mean(),
self.ratedata[:, 2].std(),
self.ratedata[:, 2].min(),
self.ratedata[:, 2].max()))
print("User Filler Mean: {}, Std: {}, Min: {}, Max: {}".format(int(O.sum(axis = 1).mean()),
int(O.sum(axis = 1).std()),
int(O.sum(axis = 1).min()),
int(O.sum(axis = 1).max())))
print("Item Filler Mean: {}, Std: {}, Min: {}, Max: {}".format(int(O.sum(axis = 0).mean()),
int(O.sum(axis = 0).std()),
int(O.sum(axis = 0).min()),
int(O.sum(axis = 0).max())))
plt.hist(O.sum(axis = 1), bins=100)
plt.xlabel("Histogram of user filler numbers")
plt.show()
plt.hist(O.sum(axis = 0), bins=100)
plt.xlabel("Histogram of item filler numbers")
plt.show()
print("User real-valued features Mean: {}, Std: {}, Min: {}, Max: {}".format(self.FeatureUser[0:self.Du].mean(axis = 1),
self.FeatureUser[0:self.Du].std(axis = 1),
self.FeatureUser[0:self.Du].min(axis = 1),
self.FeatureUser[0:self.Du].max(axis = 1)))
counts = self.FeatureUser[self.Du:].sum(axis = 1).astype(int)
print("User cat-valued feature counts:")
ind = 0
for i in self.Mu:
print(counts[ind:ind + i], self.I - counts[ind:ind + i].sum())
ind = ind + i
print("Item real-valued features Mean: {}, Std: {}, Min: {}, Max: {}".format(self.FeatureItem[0:self.Di].mean(axis = 1),
self.FeatureItem[0:self.Di].std(axis = 1),
self.FeatureItem[0:self.Di].min(axis = 1),
self.FeatureItem[0:self.Di].max(axis = 1)))
counts = self.FeatureItem[self.Di:].sum(axis = 1).astype(int)
print("Item cat-valued feature counts:")
ind = 0
for i in self.Mi:
print(counts[ind:ind + i], self.I - counts[ind:ind + i].sum())
ind = ind + i
def movie1mload(self):
filename = os.path.join(dirname, "Datasets/ml-1m/users.dat")
usrdatac = np.genfromtxt(filename, delimiter='::', usecols= [2])
usrdatad = np.genfromtxt(filename, delimiter='::', usecols= [1, 3], dtype = str)
Ddu = []
ind = np.unique(usrdatad[:,0])
Ddu.append(ind.shape[0])
for i in range(0, ind.shape[0]):
usrdatad[np.where(usrdatad[:,0] == ind[i] ), 0] = i
ind = np.unique(usrdatad[:,1])
Ddu.append(ind.shape[0])
Ddu = np.array(Ddu)
Mu = Ddu - 1
usrdatad = usrdatad.astype(int)
I = usrdatac.shape[0]
usrdatacat = np.zeros((I, sum(Ddu)-Ddu.shape[0]))
temp = np.zeros((I, Mu[0]+1))
temp[np.arange(I), usrdatad[:,0]-1] = 1
usrdatacat[:,0:Mu[0]] = temp[:,0:Mu[0]]
temp = np.zeros((I, Mu[1]+1))
temp[np.arange(I), usrdatad[:,1]-1] = 1
usrdatacat[:,Mu[0]:Mu[0] + Mu[1]] = temp[:,0:Mu[1]]
Xorg = usrdatac.T[None]
Yorg = usrdatacat.T
Mu = np.array(Mu)
Du = 1
filename = os.path.join(dirname, "Datasets/ml-1m/movies.dat")
itmdatac = np.genfromtxt(filename, delimiter='::', usecols= [1], dtype = str)
itmdatad = np.genfromtxt(filename, delimiter='::', usecols= [2], dtype = str)
itmmap = np.genfromtxt(filename, delimiter='::', usecols= [0], dtype = int)
J = itmdatac.shape[0]
for i in range(0, J):
itmdatac[i] = itmdatac[i][-5:-1]
Zorg = itmdatac.astype(float)[None]
Genres = []
for i in range(0, len(itmdatad)):
Genres = np.append(Genres, itmdatad[i].split("|"))
ind = np.unique(Genres)
Ddi = (ind.shape[0])
Mi = np.ones(Ddi).astype(int)
itmdatacat = np.zeros((Ddi, J))
for i in range(0, J):
for j in range(0, len(itmdatad[i].split("|"))):
itmdatacat[np.where(ind == itmdatad[i].split("|")[j])[0], i] = 1
Porg = itmdatacat.astype(int)
Di = 1
filename = os.path.join(dirname, "Datasets/ml-1m/ratings.dat")
ratedata = np.genfromtxt(filename, delimiter='::')
ratedata = ratedata[:,0:3]
for i in range(0, len(ratedata)):
ratedata[i, 1] = np.where(itmmap == ratedata[i,1])[0]+1
self.FeatureUser = np.append(Xorg, Yorg, axis = 0)
self.FeatureItem = np.append(Zorg, Porg, axis = 0)
self.ratedata = ratedata
self.I = I
self.J = J
self.Di = Di
self.Du = Du
self.Mi = Mi
self.Mu = Mu
def split_users(dataset, SplitSize, random, offset):
'''
Split the users into two sets and adjust rating and feature vectors accordingly
'''
ind_gen_user = np.arange(0, dataset.I)
if random == 1:
np.random.shuffle(ind_gen_user)
else:
ind_gen_user = ind_gen_user + offset
R, O = rate_to_matrix(dataset.ratedata, dataset.I, dataset.J)
R_split = R[ind_gen_user[0:SplitSize], :]
O_split = O[ind_gen_user[0:SplitSize], :]
R_rest = np.delete(R, ind_gen_user[0:SplitSize], axis = 0)
O_rest = np.delete(O, ind_gen_user[0:SplitSize], axis = 0)
dataset.ratedata = np.append(np.where(O_rest > 0)[0][None].T + 1, np.where(O_rest > 0)[1][None].T + 1, axis = 1)
dataset.ratedata = np.append(dataset.ratedata, R_rest[np.where(O_rest > 0)][None].T, axis = 1)
ratedata_split = np.append(np.where(O_split > 0)[0][None].T + 1, np.where(O_split > 0)[1][None].T + 1, axis = 1)
ratedata_split = np.append(ratedata_split, R_split[np.where(O_split > 0)][None].T, axis = 1)
FeatureUserSplit = dataset.FeatureUser[:, ind_gen_user[0:SplitSize]]
dataset.FeatureUser = np.delete(dataset.FeatureUser, ind_gen_user[0:SplitSize], axis = 1)
dataset.I = dataset.I - SplitSize
dataset_split = DatasetPrep()
dataset_split.I = SplitSize
dataset_split.J = dataset.J
dataset_split.Di = dataset.Di
dataset_split.Du = dataset.Du
dataset_split.Mi = dataset.Mi
dataset_split.Mu = dataset.Mu
dataset_split.FeatureUser = FeatureUserSplit
dataset_split.FeatureItem = dataset.FeatureItem
dataset_split.ratedata = ratedata_split
return dataset, dataset_split
def train_test_split(ratedata, InfoData, option = 'warm'):
"""
Train\Test splitting of interactions
Argument 'option' choses between 'warm' or 'cold' start scenario
"""
J = InfoData.J
I = InfoData.I
ratedata_test = np.empty((0,ratedata.shape[1]))
ratedata_train = np.empty((0,ratedata.shape[1]))
ratedata_val = np.empty((0,ratedata.shape[1]))
if option == 'warm':
for i in range(0, J):
itemdatarate = len(np.where(ratedata[:,1] == i + 1)[0])
if itemdatarate < 5:
ind = np.where(ratedata[:,1] == i+1)[0]
ratedata_train = np.append(ratedata_train, ratedata[ind, :], axis = 0)
else:
itemdatarate_test = int(float(itemdatarate) / 5)
ind = np.where(ratedata[:,1] == i+1)[0]
np.random.shuffle(ind)
ratedata_test = np.append(ratedata_test, ratedata[ind[0:itemdatarate_test], :], axis = 0)
ratedata_val = np.append(ratedata_val, ratedata[ind[itemdatarate_test:2*itemdatarate_test], :], axis = 0)
ratedata_train = np.append(ratedata_train, ratedata[ind[2*itemdatarate_test:], :], axis = 0)
elif option == 'cold_item':
indtest = np.arange(J)
np.random.shuffle(indtest)
for i in range(0, J):
ind = np.where(ratedata[:,1] == indtest[i]+1)[0]
if i < int(J/5):
ratedata_test = np.append(ratedata_test, ratedata[ind, :], axis = 0)
elif i < int(2*J/5):
ratedata_val = np.append(ratedata_val, ratedata[ind, :], axis = 0)
else:
ratedata_train = np.append(ratedata_train, ratedata[ind, :], axis = 0)
elif option == 'cold_user':
indtest = np.arange(I)
np.random.shuffle(indtest)
for i in range(0, I):
ind = np.where(ratedata[:,0] == indtest[i]+1)[0]
if i < int(I/5):
ratedata_test = np.append(ratedata_test, ratedata[ind, :], axis = 0)
elif i < int(2*I/5):
ratedata_val = np.append(ratedata_val, ratedata[ind, :], axis = 0)
else:
ratedata_train = np.append(ratedata_train, ratedata[ind, :], axis = 0)
return ratedata_train, ratedata_test, ratedata_val
def rate_to_matrix(ratedata, I, J):
R = np.zeros((I,J))
C = np.zeros((I,J), dtype = np.int8)
R[ratedata[:,0].astype(int)-1, ratedata[:,1].astype(int)-1] = ratedata[:,2]
C[ratedata[:,0].astype(int)-1, ratedata[:,1].astype(int)-1] = 1
return R, C