-
Notifications
You must be signed in to change notification settings - Fork 0
/
script.py
287 lines (254 loc) · 10.2 KB
/
script.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
#需要的程序包加载---------------------------------------------------------------------------------------------------------
from ast import literal_eval
import pandas as pd
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
from scipy import sparse as sp_sparse
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.preprocessing import MultiLabelBinarizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.multiclass import OneVsRestClassifier
from sklearn.linear_model import LogisticRegression, RidgeClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import f1_score
from sklearn.metrics import roc_auc_score
from sklearn.metrics import average_precision_score
from sklearn.metrics import recall_score
from sklearn.svm import LinearSVC
from sklearn.pipeline import Pipeline
from sklearn.feature_extraction.text import CountVectorizer
import nltk
import re
#运行之前,需要确认,电脑中的python有certificate
from nltk.corpus import stopwords
nltk.download('stopwords')
stop_words = set(stopwords.words('english'))
#数据输入----------------------------------------------------------------------------------------------------------------
df = pd .read_csv('/Users/apple/Downloads/toxic_comments-2.csv')
int(df.describe().iloc[0,0])
#数据分类与清洗-----------------------------------------------------------------------------------------------------------
train=df.iloc[range(100000),range(1,8)]
validation = df.iloc[range(100001,len(df['comment_text'])),range(1,8)]
print(len(train)); print(len(validation))
#用空格替换各种符号;删除多余符号
REPLACE_BY_SPACE_RE = re.compile('[/(){}\[\]\|@,;]')
BAD_SYMBOLS_RE = re.compile('[^0-9a-z ]')
STOPWORDS = set(stopwords.words('english'))
#数据清洗模型建立
def text_prepare(text):
text = text.lower() # 字母小写化
text = REPLACE_BY_SPACE_RE.sub(' ',text)
text = BAD_SYMBOLS_RE.sub('',text)
text = ' '.join([w for w in text.split() if w not in STOPWORDS]) # 删除停用词
return text
#创建train的数据
X_train = train.comment_text
X_val= validation.comment_text
X_train = [text_prepare(x) for x in X_train]
X_val = [text_prepare(x) for x in X_val]
#生成词频
cv = CountVectorizer(min_df=5,max_df=0.9,ngram_range=(1,2),token_pattern= '(\S+)')
feature = cv.fit_transform(X_train)
print(feature.shape); print(feature)
#词频权重
tfidf = TfidfVectorizer(min_df=5,max_df=0.9,ngram_range=(1,2),token_pattern= '(\S+)')
feature = tfidf.fit_transform(X_train)
print(feature.shape); print(feature)
#取y_train--------------------------------------------------------------------------------------------------------------
toxi=train['toxic']; toxi=list(toxi)
severe_toxic=train['severe_toxic']; severe_toxic=list(severe_toxic)
obscen=train['obscene']; obscen=list(obscen)
threat=train['threat']; threat=list(threat)
insult=train['insult']; insult=list(insult)
identity_hate=train['identity_hate']; identity_hate=list(identity_hate)
#取词与手动评
toxic = []; sever = []; obsce = []
threa = []; insul = []; ident = []
#toxic/words 创建
for item in range(len(toxi)):
a=[toxi[item]]*1
toxic.extend(a)
#severe_toxic/words 创建
for item in range(len(severe_toxic)):
a=[severe_toxic[item]]*1
sever.extend(a)
#obscene/words 创建
for item in range(len(obscen)):
a=[obscen[item]]*1
obsce.extend(a)
#threat/words 创建
for item in range(len(threat)):
a=[threat[item]]*1
threa.extend(a)
#insult/words 创建
for item in range(len(insult)):
a=[insult[item]]*1
insul.extend(a)
#identity_hate/words 创建
for item in range(len(identity_hate)):
a=[identity_hate[item]]*1
ident.extend(a)
print(len(toxic)); print(len(sever)); print(len(obsce)); print(len(threa)); print(len(insul)); print(len(ident))
a=[toxic,sever,obsce,threa,insul,ident]; y_train = []
for i in range(len(a[0])):
n = []
for j in range(len(a)):
n.append(a[j][i])
y_train.append(n)
#取y_val----------------------------------------------------------------------------------------------------------------
toxi=validation['toxic']; toxi=list(toxi)
severe_toxic=validation['severe_toxic']; severe_toxic=list(severe_toxic)
obscen=validation['obscene']; obscen=list(obscen)
threat=validation['threat']; threat=list(threat)
insult=validation['insult']; insult=list(insult)
identity_hate=validation['identity_hate']; identity_hate=list(identity_hate)
#取词与手动评
toxic = []; sever = []; obsce = []
threa = []; insul = []; ident = []
#toxic/words 创建
for item in range(len(toxi)):
a=[toxi[item]]*1
toxic.extend(a)
#severe_toxic/words 创建
for item in range(len(severe_toxic)):
a=[severe_toxic[item]]*1
sever.extend(a)
#obscene/words 创建
for item in range(len(obscen)):
a=[obscen[item]]*1
obsce.extend(a)
#threat/words 创建
for item in range(len(threat)):
a=[threat[item]]*1
threa.extend(a)
#insult/words 创建
for item in range(len(insult)):
a=[insult[item]]*1
insul.extend(a)
#identity_hate/words 创建
for item in range(len(identity_hate)):
a=[identity_hate[item]]*1
ident.extend(a)
print(len(toxic)); print(len(sever)); print(len(obsce)); print(len(threa)); print(len(insul)); print(len(ident))
a=[toxic,sever,obsce,threa,insul,ident]; y_val = []
for i in range(len(a[0])):
n = []
for j in range(len(a)):
n.append(a[j][i])
y_val.append(n)
#数据类型转换
y_train=np.array(y_train)
y_val=np.array(y_val)
#模型的建立--------------------------------------------------------------------------------------------------------------
def print_evaluation_scores(y_val, predicted):
accuracy = accuracy_score(y_val, predicted)
f1_score_macro = f1_score(y_val, predicted, average='macro')
f1_score_micro = f1_score(y_val, predicted, average='micro')
f1_score_weighted = f1_score(y_val, predicted, average='weighted')
print("accuracy:", accuracy)
print("f1_score_macro:", f1_score_macro)
print("f1_score_micro:", f1_score_micro)
print("f1_score_weighted:", f1_score_weighted)
import warnings
import sklearn.exceptions
warnings.filterwarnings("ignore", category=sklearn.exceptions.UndefinedMetricWarning)
# TF-IDF+朴素贝叶斯模型---------------------------------------------------------------------------------------------------
NB_pipeline = Pipeline([
('tfidf', TfidfVectorizer(min_df=5, max_df=0.9, ngram_range=(1, 2), token_pattern='(\S+)')),
('clf', OneVsRestClassifier(MultinomialNB())),
])
NB_pipeline.fit(X_train, y_train)
prob=NB_pipeline.predict_proba(X_val)
predicted = NB_pipeline.predict(X_val)
print_evaluation_scores(y_val, predicted)
#TF-IDF+逻辑回归---------------------------------------------------------------------------------------------------------
LogReg_pipeline = Pipeline([
('tfidf', TfidfVectorizer(min_df=5, max_df=0.9, ngram_range=(1, 2), token_pattern='(\S+)')),
('clf', OneVsRestClassifier(LogisticRegression(solver='lbfgs',max_iter=10000), n_jobs=1)),
])
LogReg_pipeline.fit(X_train, y_train)
predicted = LogReg_pipeline.predict(X_val)
print_evaluation_scores(y_val, predicted)
#CountVectorizer+朴素贝叶斯----------------------------------------------------------------------------------------------
NB_pipeline = Pipeline([
('cv', CountVectorizer(min_df=5, max_df=0.9, ngram_range=(1, 2), token_pattern='(\S+)')),
('clf', OneVsRestClassifier(MultinomialNB(
fit_prior=True, class_prior=None))),
])
NB_pipeline.fit(X_train, y_train)
predicted = NB_pipeline.predict(X_val)
print_evaluation_scores(y_val, predicted)
#CountVectorizer+逻辑回归------------------------------------------------------------------------------------------------
LogReg_pipeline = Pipeline([
('cv', CountVectorizer(min_df=5, max_df=0.9, ngram_range=(1, 2), token_pattern='(\S+)')),
('clf', OneVsRestClassifier(LogisticRegression(solver='lbfgs',max_iter=10000), n_jobs=1)),
])
LogReg_pipeline.fit(X_train, y_train)
predicted = LogReg_pipeline.predict(X_val)
print_evaluation_scores(y_val, predicted)
# 模型预测--返回每个测试评论为toxic的概率-----------------------------------------------------------------------------------
#模型建立
LogReg_pipeline = Pipeline([
('tfidf', TfidfVectorizer(min_df=5, max_df=0.9, ngram_range=(1, 2), token_pattern='(\S+)')),
('clf', OneVsRestClassifier(LogisticRegression(solver='lbfgs',max_iter=10000), n_jobs=1)),
])
#取y_total整体数据
toxi=df['toxic']; toxi=list(toxi)
severe_toxic=df['severe_toxic']; severe_toxic=list(severe_toxic)
obscen=df['obscene']; obscen=list(obscen)
threat=df['threat']; threat=list(threat)
insult=df['insult']; insult=list(insult)
identity_hate=df['identity_hate']; identity_hate=list(identity_hate)
#取词与手动评
toxic = []; sever = []; obsce = []
threa = []; insul = []; ident = []
#toxic/words 创建
for item in range(len(toxi)):
a=[toxi[item]]*1
toxic.extend(a)
#severe_toxic/words 创建
for item in range(len(severe_toxic)):
a=[severe_toxic[item]]*1
sever.extend(a)
#obscene/words 创建
for item in range(len(obscen)):
a=[obscen[item]]*1
obsce.extend(a)
#threat/words 创建
for item in range(len(threat)):
a=[threat[item]]*1
threa.extend(a)
#insult/words 创建
for item in range(len(insult)):
a=[insult[item]]*1
insul.extend(a)
#identity_hate/words 创建
for item in range(len(identity_hate)):
a=[identity_hate[item]]*1
ident.extend(a)
print(len(toxic)); print(len(sever)); print(len(obsce)); print(len(threa)); print(len(insul)); print(len(ident))
a=[toxic,sever,obsce,threa,insul,ident]; y_total = []
for i in range(len(a[0])):
n = []
for j in range(len(a)):
n.append(a[j][i])
y_total.append(n)
y_total=np.array(y_total)
#取X-total
X_total = df.comment_text
X_total = [text_prepare(x) for x in X_total]
#预测
LogReg_pipeline.fit(X_total, y_total)
predicted = LogReg_pipeline.predict(X_total)
score=LogReg_pipeline.predict_proba(X_total)
#输出
final=[]
for i in range(len(y_total)):
finl=[y_total[i],score[i]]
final.extend(finl)
final=np.array(final)
name=['toxic','severe_toxic','obscene','threat','insult','identity_hate']
prediction=pd.DataFrame(columns=name,data=final)
prediction.to_csv('/Users/apple/Desktop/prediction.csv')