-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathregression.py
190 lines (143 loc) · 6.85 KB
/
regression.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
import data_handling
import numpy as np
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfTransformer, TfidfVectorizer
from sklearn.linear_model import SGDClassifier, SGDRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
from sklearn.svm import LinearSVC, SVR
from evaluation import rmslog_error
class BaseBowRegressor(object):
# Items can be trained for different outputs
FUNNY_VOTES = 0
COOL_VOTES = 1
USEFUL_VOTES = 2
STARS = 3
def __init__(self, ngram_range=(1,1)):
self.ngram_range = ngram_range
# Labels are given in groups since we can train the system using different
# outputs. Labels, test labels and regressors are indexed according to the
# categories given above
self.reviews = []
self.labels = [[], [], [], []]
self.test_reviews = []
self.test_labels = [[], [], [], []]
self.regs = [None, None, None, None]
@staticmethod
def get_reviews_data(partitions_to_use):
"""
Gets data from reviews and returns it so that the class can use it to load
training or test data
"""
data = data_handling.load_partitions(partitions_to_use)
review_texts = []
useful_votes = []
funny_votes = []
cool_votes = []
review_stars = []
for review in data:
review_texts.append(review['text'])
useful_votes.append(review['votes']['useful'])
cool_votes.append(review['votes']['cool'])
funny_votes.append(review['votes']['funny'])
review_stars.append(review['stars'])
return review_texts, useful_votes, funny_votes, cool_votes, review_stars
def load_training_data(self, partitions_to_use):
self.reviews, useful, funny, cool, stars =\
self.get_reviews_data(partitions_to_use)
self.labels[self.USEFUL_VOTES] = useful
self.labels[self.FUNNY_VOTES] = funny
self.labels[self.COOL_VOTES] = cool
self.labels[self.STARS] = stars
def load_test_data(self, partitions_to_use):
"""
Loads test data into the object
"""
self.test_reviews, test_useful, test_funny, test_cool, test_stars =\
self.get_reviews_data(partitions_to_use)
self.test_labels[self.USEFUL_VOTES] = test_useful
self.test_labels[self.FUNNY_VOTES] = test_funny
self.test_labels[self.COOL_VOTES] = test_cool
self.test_labels[self.STARS] = test_stars
def train(self):
raise NotImplementedError()
def __test(self, reviews, labels):
raise NotImplementedError()
def get_bag_of_ngrams(self, texts, ngram_range=None):
""" Sets vectorizer feature and returns data from object in feature form X """
if ngram_range is None:
ngram_range = self.ngram_range
self.count_vect = CountVectorizer(ngram_range=ngram_range, stop_words="english")
X_train_counts = self.count_vect.fit_transform(texts)
self.tfidf_transformer = TfidfTransformer(use_idf=True).fit(X_train_counts)
X_train_tfidf = self.tfidf_transformer.transform(X_train_counts)
return X_train_tfidf
DEFAULT_LABEL = BaseBowRegressor.FUNNY_VOTES
class SGD(BaseBowRegressor):
"""
Stochastic Gradient Descent with Tfidf
"""
def train(self, train_on=DEFAULT_LABEL, limit_data=None):
if not hasattr(self, 'reviews'):
print "No data loaded"
return
if limit_data is None:
limit_data = len(self.reviews)
X = self.get_bag_of_ngrams(self.reviews[:limit_data])
self.regs[train_on] = SGDRegressor(loss="huber", alpha=0.0001, penalty="l1", n_iter=20).fit(X, self.labels[train_on][:limit_data])
def __test(self, reviews, labels, test_on=DEFAULT_LABEL):
X_training_counts = self.count_vect.transform(reviews)
X_training_tfidf = self.tfidf_transformer.transform(X_training_counts)
predicted = self.regs[test_on].predict(X_training_tfidf)
return rmslog_error(predicted, labels), rmslog_error(np.zeros(len(predicted)), labels)
def get_training_error(self, train_on=DEFAULT_LABEL):
return self.__test(self.reviews, self.labels[train_on])
def get_generalized_error(self, test_on=DEFAULT_LABEL):
return self.__test(self.test_reviews, self.test_labels[test_on])
def __get_scores(self, reviews, labels, train_on=DEFAULT_LABEL):
X_training_counts = self.count_vect.transform(reviews)
X_training_tfidf = self.tfidf_transformer.transform(X_training_counts)
return self.regs[train_on].score(X_training_tfidf, labels)
def get_training_R2(self, train_on=DEFAULT_LABEL):
return self.__get_scores(self.reviews, self.labels[train_on], train_on)
def get_test_R2(self, test_on=DEFAULT_LABEL):
return self.__get_scores(self.test_reviews, self.test_labels[test_on], test_on)
class SupportVectorRegressor(BaseBowRegressor):
"""
Stochastic Gradient Descent with Tfidf
"""
def train(self, train_on=DEFAULT_LABEL, limit_data=None):
if not hasattr(self, 'reviews'):
print "No data loaded"
return
if limit_data is None:
limit_data = len(self.reviews)
X = self.get_bag_of_ngrams(self.reviews[:limit_data])
self.regs[train_on] = SVR(kernel='rbf').fit(X, self.labels[train_on][:limit_data])
def __test(self, reviews, labels, test_on=DEFAULT_LABEL):
X_training_counts = self.count_vect.transform(reviews)
X_training_tfidf = self.tfidf_transformer.transform(X_training_counts)
predicted = self.regs[test_on].predict(X_training_tfidf)
return rmslog_error(predicted, labels), rmslog_error(np.zeros(len(predicted)), labels)
def get_training_error(self, train_on=DEFAULT_LABEL):
return self.__test(self.reviews, self.labels[train_on])
def get_generalized_error(self, test_on=DEFAULT_LABEL):
return self.__test(self.test_reviews, self.test_labels[test_on])
def __get_scores(self, reviews, labels, train_on=DEFAULT_LABEL):
X_training_counts = self.count_vect.transform(reviews)
X_training_tfidf = self.tfidf_transformer.transform(X_training_counts)
return self.regs[train_on].score(X_training_tfidf, labels)
def get_training_R2(self, train_on=DEFAULT_LABEL):
return self.__get_scores(self.reviews, self.labels[train_on], train_on)
def get_test_R2(self, test_on=DEFAULT_LABEL):
return self.__get_scores(self.test_reviews, self.test_labels[test_on], test_on)
if __name__ == "__main__":
# Examples
#sgd = SGD()
sgd = SupportVectorRegressor()
sgd.load_training_data(range(1, 2))
sgd.load_test_data(range(3,4))
sgd.train()
print sgd.get_training_error()
print sgd.get_generalized_error()
print sgd.get_training_R2()
print sgd.get_test_R2()