This repository has been archived by the owner on Nov 24, 2017. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathencoder.py
307 lines (249 loc) · 8.13 KB
/
encoder.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
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
# Author: Hongwei Zhang
# Email: [email protected]
import mmh3
import numpy as np
from scipy.sparse import coo_matrix
from abc import ABC, abstractmethod
class Encoder(ABC):
@abstractmethod
def fit(self, X):
pass
@abstractmethod
def transform(self, X):
pass
class FFMHashEncoder(Encoder):
def __init__(self, hash_base=1000000, hash_offset=100, seed=0):
self._counter = dict()
self._hash_base = hash_base
self._hash_offset = hash_offset
self._seed = seed
def fit(self, X):
"""
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
Returns
-------
self
"""
for x in X:
for word in x:
try:
self._counter[word] += 1
except KeyError:
self._counter[word] = 1
return self
def transform(self, X, threshold=20, dummy_field=False):
"""
unpresented word will be dropped
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
threshold: words that occur less than threshold will be dropped
dummy_field: in order to follow input format of ffm,
use the same dummy field for all features
Returns
-------
FFM format data: [[filed1:hash(filed1_index1):1,
filed1:hash(filed1_index2):1,
filed2:hash(filed2_index1):1],
[filed1:hash(field1_index3):1,
filed3:hash(field3_index1):1]
"""
result = list()
for x in X:
row_result = list()
for word in x:
if word in self._counter and self._counter[word] >= threshold:
field = "0"
if dummy_field is False:
field = word.split("_")[0]
index = str(
mmh3.hash(
word,
self._seed, signed=False) % self._hash_base +
self._hash_offset)
row_result.append(":".join([field, index, "1"]))
result.append(row_result)
return result
class OneHotEncoder(Encoder):
def __init__(self):
self._encoder = dict()
self._counter = dict()
def fit(self, X, threshold=3):
"""
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
threshold: words that occur less than threshold will be dropped
Returns
-------
self
"""
for x in X:
for word in x:
try:
self._counter[word] += 1
except KeyError:
self._counter[word] = 1
first_index = 0
for key, value in self._counter.items():
if value >= threshold:
self._encoder[key] = first_index
first_index += 1
return self
def transform(self, X):
"""
unpresented word will be dropped
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
Returns
-------
csr matrix
"""
nrows = len(X)
ncols = len(self._encoder)
row_indices = list()
col_indices = list()
data = list()
for row in range(nrows):
x = X[row]
for word in x:
if word in self._encoder:
col = self._encoder[word]
row_indices.append(row)
col_indices.append(col)
data.append(1)
return coo_matrix((
np.array(data),
(np.array(row_indices), np.array(col_indices))),
shape=(nrows, ncols)).tocsr()
@property
def codebook(self):
return self._encoder
class SVMHashEncoder(Encoder):
def __init__(self, hash_base=100000, hash_offset=100, seed=0):
self._counter = dict()
self._hash_base = hash_base
self._hash_offset = hash_offset
self._seed = seed
def fit(self, X):
"""
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
Returns
-------
self
"""
for x in X:
for word in x:
try:
self._counter[word] += 1
except KeyError:
self._counter[word] = 1
return self
def transform(self, X, threshold=3):
"""
unpresented word will be dropped
index is not in ascending order!!!
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
threshold: words that occur less than threshold will be dropped
Returns
-------
LibSVM format data: [[hash(filed1_index1):1,
hash(filed1_index2):1,
hash(filed2_index1):1],
[hash(filed1_index3):1,
hash(filed3_index1):1]]
"""
result = list()
for x in X:
row_result = list()
for word in x:
if word in self._counter and self._counter[word] >= threshold:
index = str(mmh3.hash(
word, self._seed, signed=False) % self._hash_base +
self._hash_offset)
row_result.append(":".join([index, "1"]))
result.append(row_result)
return result
class BinaryStatsCalculator(Encoder):
def __init__(self):
self._encoder = dict()
self._counter = dict()
def fit(self, X, y, threshold=3):
"""
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
y: [0, 1]
threshold: words that occur less than threshold will be dropped
Returns
-------
self
"""
for x in X:
for word in x:
try:
self._counter[word] += 1
except KeyError:
self._counter[word] = 1
first_index = 0
for key, value in self._counter.items():
if value >= threshold:
self._encoder[key] = [first_index, 0, 0]
first_index += 1
for row in range(len(X)):
words = X[row]
label = y[row]
for word in words:
if word in self._encoder:
self._encoder[word][1] += 1
if label == 1:
self._encoder[word][2] += 1
return self
def transform(self, X):
"""
unpresented word will be dropped
index is not in ascending order!!!
Parameters
----------
X: [[field1_index1, filed1_index2, field2_index1],
[field1_index3, field3_index1]]
Returns
-------
csr matrix
"""
nrows = len(X)
ncols = len(self._encoder)
row_indices = list()
col_indices = list()
data = list()
for row in range(nrows):
x = X[row]
for word in x:
if word in self._encoder:
col = self._encoder[word][0]
row_indices.append(row)
col_indices.append(col)
data.append(
1.0 * self._encoder[word][2] /
self._encoder[word][1])
return coo_matrix((
np.array(data),
(np.array(row_indices), np.array(col_indices))),
shape=(nrows, ncols)).tocsr()
@property
def codebook(self):
return self._encoder