-
Notifications
You must be signed in to change notification settings - Fork 2
/
depression_detection.py
418 lines (345 loc) · 15.3 KB
/
depression_detection.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
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
# -*- coding: utf-8 -*-
"""Depression Detection.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1gJah95H-y_qO-_ldBzI8nA0uXQaADB3r
"""
from google.colab import drive
drive.mount('/content/drive')
# Commented out IPython magic to ensure Python compatibility.
!pwd
!ls
# %cd /content/drive/MyDrive/depression
import pandas as pd
#pandas
import numpy as np
#numpy
import matplotlib.pyplot as plt
#matplotlib
import seaborn as sns
#seaborn
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from sklearn.feature_extraction.text import CountVectorizer
#sklearn
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.models import Sequential
from keras.layers import Activation, Dense, Dropout, Embedding, Flatten, Conv1D, MaxPooling1D, LSTM
#from tensorflow.keras.layers import Conv2D, MaxPooling2D
from tensorflow.keras.layers import BatchNormalization
from keras import utils
from keras.callbacks import ReduceLROnPlateau, EarlyStopping
from keras.optimizers import Adam
#keras
import tensorflow as tf
#tensorflow
import nltk
nltk.download('stopwords')
from nltk.corpus import stopwords
from nltk.stem import SnowballStemmer
#nltk
import os
import nltk # For NLP
import re # For Regex
import string # For punctuation
#other useful stuff
from wordcloud import WordCloud, STOPWORDS
#wordclouds and cloud stopwords
import warnings
warnings.filterwarnings("ignore")
data = pd.read_csv('training.1600000.processed.noemoticon.csv', encoding = 'latin', header=None)
data.head(10)
data = data.rename(columns={0: 'target', 1: 'id', 2: 'date', 3: 'query', 4: 'username', 5: 'content'})
data.head(10)
missing_data = data.isna().sum().sort_values(ascending=False)
percentage_missing = round((data.isnull().sum()/data.isnull().count()).sort_values(ascending=False)*100,2)
missing_info = pd.concat([missing_data,percentage_missing],keys=['Missing values','Percentage'],axis=1)
missing_info.style.background_gradient()
pd.set_option('display.max_colwidth', -1)
data[data['target']==0]['content'].head()
data[data['target']==4]['content'].head
data[data['target']==2]['content'].head
data['target'] = data['target'].replace([0, 4],['Negative','Positive'])
fig = plt.figure(figsize=(8,8))
targets = data.groupby('target').size()
targets.plot(kind='pie', subplots=True, figsize=(10, 8), autopct = "%.2f%%", colors=['red','green'])
plt.title("Pie chart of different classes of tweets",fontsize=16)
plt.ylabel("")
plt.legend()
plt.show()
data['target'].value_counts()
data['length'] = data.content.str.split().apply(len)
fig = plt.figure(figsize=(14,7))
ax1 = fig.add_subplot(122)
sns.distplot(data[data['target']=='Positive']['length'], ax=ax1,color='green')
describe = data.length[data.target=='Positive'].describe().to_frame().round(2)
ax2 = fig.add_subplot(121)
ax2.axis('off')
font_size = 14
bbox = [0, 0, 1, 1]
table = ax2.table(cellText = describe.values, rowLabels = describe.index, bbox=bbox, colLabels=describe.columns)
table.set_fontsize(font_size)
fig.suptitle('Distribution of text length for positive sentiment tweets.', fontsize=16)
plt.show()
fig = plt.figure(figsize=(14,7))
ax1 = fig.add_subplot(122)
sns.distplot(data[data['target']=='Negative']['length'], ax=ax1,color='red')
describe = data.length[data.target=='Negative'].describe().to_frame().round(2)
ax2 = fig.add_subplot(121)
ax2.axis('off')
font_size = 14
bbox = [0, 0, 1, 1]
table = ax2.table(cellText = describe.values, rowLabels = describe.index, bbox=bbox, colLabels=describe.columns)
table.set_fontsize(font_size)
fig.suptitle('Distribution of text length for negative sentiment tweets.', fontsize=16)
plt.show()
plt.figure(figsize=(14,7))
common_keyword=sns.barplot(x=data[data['target']=='Positive']['username'].value_counts()[:10].index, \
y=data[data['target']=='Positive']['username'].value_counts()[:10],palette='viridis')
common_keyword.set_xticklabels(common_keyword.get_xticklabels(),rotation=90)
common_keyword.set_ylabel('Positive tweet frequency',fontsize=12)
plt.title('Top 10 users who publish positive tweets',fontsize=16)
plt.show()
plt.figure(figsize=(14,7))
common_keyword=sns.barplot(x=data[data['target']=='Negative']['username'].value_counts()[:10].index, \
y=data[data['target']=='Negative']['username'].value_counts()[:10],palette='Spectral')
common_keyword.set_xticklabels(common_keyword.get_xticklabels(),rotation=90)
common_keyword.set_ylabel('Negative tweet frequency',fontsize=12)
plt.title('Top 10 users who publish negative tweets',fontsize=16)
plt.show()
plt.figure(figsize=(14,7))
word_cloud = WordCloud(stopwords = STOPWORDS, max_words = 200, width=1366, height=768, background_color="white").generate(" ".join(data[data.target=='Positive'].content))
plt.imshow(word_cloud,interpolation='bilinear')
plt.axis('off')
plt.title('Most common words in positive sentiment tweets.',fontsize=20)
plt.show()
plt.figure(figsize=(14,7))
word_cloud = WordCloud(stopwords = STOPWORDS, max_words = 200, width=1366, height=768, background_color="white").generate(" ".join(data[data.target=='Negative'].content))
plt.imshow(word_cloud,interpolation='bilinear')
plt.axis('off')
plt.title('Most common words in negative sentiment tweets.',fontsize=20)
plt.show()
data.drop(['id','date','query','username','length'], axis=1, inplace=True)
data.target = data.target.replace({'Positive': 1, 'Negative': 0})
from string import punctuation
print("DATA CLEANING -- \n")
# emojis defined
emoji_pattern = re.compile("["
u"\U0001F300-\U0001F5FF" # symbols & pictographs
u"\U0001F680-\U0001F6FF" # transport & map symbols
u"\U0001F1E0-\U0001F1FF" # flags (iOS)
u"\U00002702-\U000027B0"
u"\U000024C2-\U0001F251"
"]+", flags=re.UNICODE)
def replace_emojis(t):
'''
This function replaces happy unicode emojis with "happy" and sad unicode emojis with "sad.
'''
emoji_happy = ["\U0001F600", "\U0001F601", "\U0001F602","\U0001F603","\U0001F604","\U0001F605", "\U0001F606", "\U0001F607", "\U0001F609",
"\U0001F60A", "\U0001F642","\U0001F643","\U0001F923",r"\U0001F970","\U0001F60D", r"\U0001F929","\U0001F618","\U0001F617",
r"\U000263A", "\U0001F61A", "\U0001F619", r"\U0001F972", "\U0001F60B", "\U0001F61B", "\U0001F61C", r"\U0001F92A",
"\U0001F61D", "\U0001F911", "\U0001F917", r"\U0001F92D", r"\U0001F92B","\U0001F914","\U0001F910", r"\U0001F928", "\U0001F610", "\U0001F611",
"\U0001F636", "\U0001F60F","\U0001F612", "\U0001F644","\U0001F62C","\U0001F925","\U0001F60C","\U0001F614","\U0001F62A",
"\U0001F924","\U0001F634", "\U0001F920", r"\U0001F973", r"\U0001F978","\U0001F60E","\U0001F913", r"\U0001F9D0"]
emoji_sad = ["\U0001F637","\U0001F912","\U0001F915","\U0001F922", r"\U0001F92E","\U0001F927", r"\U0001F975", r"\U0001F976", r"\U0001F974",
"\U0001F635", r"\U0001F92F", "\U0001F615","\U0001F61F","\U0001F641", r"\U0002639","\U0001F62E","\U0001F62F","\U0001F632",
"\U0001F633", r"\U0001F97A","\U0001F626","\U0001F627","\U0001F628","\U0001F630","\U0001F625","\U0001F622","\U0001F62D",
"\U0001F631","\U0001F616","\U0001F623" ,"\U0001F61E","\U0001F613","\U0001F629","\U0001F62B", r"\U0001F971",
"\U0001F624","\U0001F621","\U0001F620", r"\U0001F92C","\U0001F608","\U0001F47F","\U0001F480", r"\U0002620"]
words = t.split()
reformed = []
for w in words:
if w in emoji_happy:
reformed.append("happy")
elif w in emoji_sad:
reformed.append("sad")
else:
reformed.append(w)
t = " ".join(reformed)
return t
def replace_smileys(t):
'''
This function replaces happy smileys with "happy" and sad smileys with "sad.
'''
emoticons_happy = set([':-)', ':)', ';)', ':o)', ':]', ':3', ':c)', ':>', '=]', '8)', '=)', ':}', ':D',
':^)', ':-D', ':D', '8-D', '8D', 'x-D', 'xD', 'X-D', 'XD', '=-D', '=D',
'=-3', '=3', ':-))', ":'-)", ":')", ':*', ':^*', '>:P', ':-P', ':P', 'X-P',
'x-p', 'xp', 'XP', ':-p', ':p', '=p', ':-b', ':b', '>:)', '>;)', '>:-)', '<3'])
emoticons_sad = set([':L', ':-/', '>:/', ':S', '>:[', ':@', ':-(', ':[', ':-||', '=L', ':<',
':-[', ':-<', '=\\', '=/', '>:(', ':(', '>.<', ":'-(", ":'(", ':\\', ':-c',
':c', ':{', '>:\\', ';('])
words = t.split()
reformed = []
for w in words:
if w in emoticons_happy:
reformed.append("happy")
elif w in emoticons_sad:
reformed.append("sad")
else:
reformed.append(w)
t = " ".join(reformed)
return t
def replace_contractions(t):
'''
This function replaces english lanuage contractions like "shouldn't" with "should not"
'''
cont = {"aren't" : 'are not', "can't" : 'cannot', "couln't": 'could not', "didn't": 'did not', "doesn't" : 'does not',
"hadn't": 'had not', "haven't": 'have not', "he's" : 'he is', "she's" : 'she is', "he'll" : "he will",
"she'll" : 'she will',"he'd": "he would", "she'd":"she would", "here's" : "here is",
"i'm" : 'i am', "i've" : "i have", "i'll" : "i will", "i'd" : "i would", "isn't": "is not",
"it's" : "it is", "it'll": "it will", "mustn't" : "must not", "shouldn't" : "should not", "that's" : "that is",
"there's" : "there is", "they're" : "they are", "they've" : "they have", "they'll" : "they will",
"they'd" : "they would", "wasn't" : "was not", "we're": "we are", "we've":"we have", "we'll": "we will",
"we'd" : "we would", "weren't" : "were not", "what's" : "what is", "where's" : "where is", "who's": "who is",
"who'll" :"who will", "won't":"will not", "wouldn't" : "would not", "you're": "you are", "you've":"you have",
"you'll" : "you will", "you'd" : "you would", "mayn't" : "may not"}
words = t.split()
reformed = []
for w in words:
if w in cont:
reformed.append(cont[w])
else:
reformed.append(w)
t = " ".join(reformed)
return t
def remove_single_letter_words(t):
'''
This function removes words that are single characters
'''
words = t.split()
reformed = []
for w in words:
if len(w) > 1:
reformed.append(w)
t = " ".join(reformed)
return t
print("Cleaning the tweets from the data.\n")
print("Replacing handwritten emojis with their feeling associated.")
print("Convert to lowercase.")
print("Replace contractions.")
print("Replace unicode emojis with their feeling associated.")
print("Remove all other unicoded emojis.")
print("Remove NON- ASCII characters.")
print("Remove numbers.")
print("Remove \"#\". ")
print("Remove \"@\". ")
print("Remove usernames.")
print("Remove \'RT\'. ")
print("Replace all URLs and Links with word \'URL\'.")
print("Remove all punctuations.")
print("Removes single letter words.\n")
def dataclean(t):
'''
This function cleans the tweets.
'''
t = replace_smileys(t) # replace handwritten emojis with their feeling associated
t = t.lower() # convert to lowercase
t = replace_contractions(t) # replace short forms used in english with their actual words
t = replace_emojis(t) # replace unicode emojis with their feeling associated
t = emoji_pattern.sub(r'', t) # remove emojis other than smiley emojis
t = re.sub('\\\\u[0-9A-Fa-f]{4}','', t) # remove NON- ASCII characters
t = re.sub("[0-9]", "", t) # remove numbers # re.sub("\d+", "", t)
t = re.sub('#', '', t) # remove '#'
t = re.sub('@[A-Za-z0–9]+', '', t) # remove '@'
t = re.sub('@[^\s]+', '', t) # remove usernames
t = re.sub('RT[\s]+', '', t) # remove retweet 'RT'
t = re.sub('((www\.[^\s]+)|(https?://[^\s]+))', '', t) # remove links (URLs/ links)
t = re.sub('[!"$%&\'()*+,-./:@;<=>?[\\]^_`{|}~]', '', t) # remove punctuations
t = t.replace('\\\\', '')
t = t.replace('\\', '')
t = remove_single_letter_words(t) # removes single letter words
return t
data['content'] = data['content'].apply(dataclean)
print("Tweets have been cleaned.")
data.head()
english_stopwords = stopwords.words('english')
#base of english stopwords
stemmer = SnowballStemmer('english')
#stemming algorithm
regex = "@\S+|https?:\S+|http?:\S|[^A-Za-z0-9]+"
#regex for mentions and links in tweets
def preprocess(content, stem=False):
content = re.sub(regex, ' ', str(content).lower()).strip()
tokens = []
for token in content.split():
if token not in english_stopwords:
tokens.append(stemmer.stem(token))
return " ".join(tokens)
data.content = data.content.apply(lambda x: preprocess(x))
data.head()
train, test = train_test_split(data, test_size=0.1, random_state=44)
print('Train dataset shape: {}'.format(train.shape))
print('Test dataset shape: {}'.format(test.shape))
tokenizer = Tokenizer()
tokenizer.fit_on_texts(train.content)
vocab_size = len(tokenizer.word_index) + 1
max_length = 50
sequences_train = tokenizer.texts_to_sequences(train.content)
sequences_test = tokenizer.texts_to_sequences(test.content)
X_train = pad_sequences(sequences_train, maxlen=max_length, padding='post')
X_test = pad_sequences(sequences_test, maxlen=max_length, padding='post')
y_train = train.target.values
y_test = test.target.values
embeddings_dictionary = dict()
embedding_dim = 100
glove_file = open('glove.6B.100d.txt')
for line in glove_file:
records = line.split()
word = records[0]
vector_dimensions = np.asarray(records[1:], dtype='float32')
embeddings_dictionary [word] = vector_dimensions
glove_file.close()
embeddings_matrix = np.zeros((vocab_size, embedding_dim))
for word, index in tokenizer.word_index.items():
embedding_vector = embeddings_dictionary.get(word)
if embedding_vector is not None:
embeddings_matrix[index] = embedding_vector
embedding_layer = tf.keras.layers.Embedding(vocab_size, embedding_dim, input_length=max_length, weights=[embeddings_matrix], trainable=False)
num_epochs = 70
batch_size = 1000
model = Sequential([
embedding_layer,
tf.keras.layers.Bidirectional(LSTM(128, return_sequences=True)),
tf.keras.layers.Dropout(0.4),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.Bidirectional(LSTM(128)),
tf.keras.layers.Dropout(0.4),
# tf.keras.layers.BatchNormalization(),
tf.keras.layers.Dense(64, activation='relu'),
tf.keras.layers.Dense(1, activation='sigmoid'),
])
model.summary()
tf.keras.utils.plot_model(model, show_shapes=True)
model.compile(loss='binary_crossentropy', optimizer=Adam(learning_rate=0.001), metrics=['accuracy'])
#es = EarlyStopping(monitor='val_loss', mode='min', verbose=1)
history = model.fit(X_train, y_train, batch_size = batch_size, epochs=num_epochs, validation_data=(X_test, y_test), verbose=1)# , callbacks=[es])
y_pred = model.predict(X_test)
y_pred = np.where(y_pred>0.5, 1, 0)
print(classification_report(y_test, y_pred))
#History for accuracy
plt.figure(figsize=(10,5))
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['Train accuracy', 'Test accuracy'], loc='lower right')
plt.show()
# History for loss
plt.figure(figsize=(10,5))
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['Train loss', 'Test loss'], loc='upper right')
plt.suptitle('Accuracy and loss for second model')
plt.show()
vec = CountVectorizer()
xeval=["I am love you "]
print(xeval)
xeval_numeric = vec.fit_transform(xeval).toarray()
prediction=model.predict(xeval_numeric)
print(prediction)
model.save("my_model2.h5")