Fake News Detection

import pandas as pd
import numpy as np
import nltk
import re

from tensorflow.keras.layers import Embedding,LSTM,Dense,Dropout
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.preprocessing.text import one_hot

# data load
df = pd.read_csv('./train.csv')
test = pd.read_csv('./test.csv')

# filling NULL values with '' string
df = df.fillna('')
test = test.fillna('')

df['total'] = df['title'] + df['author']
test['total'] = test['title'] + test['author']

X = df.drop('label', axis=1)
y = df['label']

#Choosing vocabulary size to be 5000 and copying data to msg for further cleaning
voc_size = 5000
msg = X.copy()
msg_test = test.copy()

from nltk.corpus import stopwords
nltk.download('stopwords')

from nltk.stem.porter import PorterStemmer
ps = PorterStemmer()
corpus = []

#Applying stemming and some preprocessing
for i in range(len(msg)):
    review = re.sub('[^a-zA-Z]', ' ', msg['total'][i])
    review = review.lower()
    review = review.split()
    review = [ps.stem(word) for word in review if not word in stopwords.words('english')]
    review = ' '.join(review)
    corpus.append(review)

corpus_test = []
for i in range(len(msg_test)):
    review = re.sub('[^a-zA-Z]', ' ', msg_test['total'][i])
    review = review.lower()
    review = review.split()
    review = [ps.stem(word) for word in review if not word in stopwords.words('english')]
    review = ' '.join(review)
    corpus_test.append(review)

# Converting to one hot representation
onehot_rep = [one_hot(words, voc_size) for words in corpus]
onehot_rep_test = [one_hot(words, voc_size) for words in corpus_test]

#Padding Sentences to make them of same size
embedded_docs = pad_sequences(onehot_rep, padding='pre', maxlen=25)
embedded_docs_test = pad_sequences(onehot_rep_test, padding='pre', maxlen=25)

#We have used embedding layers with LSTM
model = Sequential()
model.add(Embedding(voc_size, 40, input_length=25))
model.add(Dropout(0.3))
model.add(LSTM(100))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
print(model.summary())
'''
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
embedding (Embedding)        (None, 25, 40)            200000    
_________________________________________________________________
dropout (Dropout)            (None, 25, 40)            0
_________________________________________________________________
lstm (LSTM)                  (None, 100)               56400
_________________________________________________________________
dense (Dense)                (None, 64)                6464
_________________________________________________________________
dropout_1 (Dropout)          (None, 64)                0
_________________________________________________________________
dense_1 (Dense)              (None, 1)                 65
=================================================================
Total params: 262,929
Trainable params: 262,929
Non-trainable params: 0
_________________________________________________________________
'''

X_final = np.array(embedded_docs)
y_final = np.array(y)
test_final = np.array(embedded_docs_test)

model.fit(X_final, y_final, epochs=20, batch_size=64)

# y_pred = model.predict_classes(test_final)
y_pred = model.predict(test_final)
print(y_pred)
y_pred = (y_pred > 0.5).astype('int32')

final_sub = pd.DataFrame()
final_sub['id'] = test['id']
final_sub['label'] = y_pred
# final_sub.to_csv('final_sub.csv', index=False)
print(final_sub.head())