!pip install -q nltk scikit-learn matplotlib seaborn wordcloud

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from wordcloud import WordCloud

from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score, roc_curve

import nltk
nltk.download('stopwords')
nltk.download('punkt')
nltk.download('wordnet')
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from nltk.stem import WordNetLemmatizer

[nltk_data] Downloading package stopwords to /root/nltk_data...
[nltk_data]   Unzipping corpora/stopwords.zip.
[nltk_data] Downloading package punkt to /root/nltk_data...
[nltk_data]   Unzipping tokenizers/punkt.zip.
[nltk_data] Downloading package wordnet to /root/nltk_data...

from google.colab import files
uploaded = files.upload()

Saving Fake.csv to Fake.csv
Saving True.csv to True.csv

import pandas as pd

fake_df = pd.read_csv("Fake.csv")
real_df = pd.read_csv("True.csv")


fake_df['label'] = 0  # Fake
real_df['label'] = 1  # Real

df = pd.concat([fake_df, real_df]).reset_index(drop=True)

df = df.sample(frac=1, random_state=42).reset_index(drop=True)

df.head()

# Drop date and subject columns if present
df = df[['title', 'text', 'label']]
df.dropna(inplace=True)  # Remove rows with missing text/title

# Combine title and text
df['content'] = df['title'] + " " + df['text']
df = df[['content', 'label']]
df.head()

<ipython-input-28-3bbdd87b4539>:3: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  df.dropna(inplace=True)  # Remove rows with missing text/title

!pip install nltk --quiet

import re
import nltk
import string

nltk.download('stopwords')
nltk.download('punkt')
nltk.download('wordnet')
nltk.download('punkt_tab')

from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer

[nltk_data] Downloading package stopwords to /root/nltk_data...
[nltk_data]   Package stopwords is already up-to-date!
[nltk_data] Downloading package punkt to /root/nltk_data...
[nltk_data]   Package punkt is already up-to-date!
[nltk_data] Downloading package wordnet to /root/nltk_data...
[nltk_data]   Package wordnet is already up-to-date!
[nltk_data] Downloading package punkt_tab to /root/nltk_data...
[nltk_data]   Unzipping tokenizers/punkt_tab.zip.

stop_words = set(stopwords.words('english'))
lemmatizer = WordNetLemmatizer()

def clean_text(text):
    # Lowercase
    text = text.lower()
    # Remove punctuation
    text = re.sub(f"[{re.escape(string.punctuation)}]", "", text)
    # Tokenize
    words = nltk.word_tokenize(text)
    # Remove stopwords and lemmatize
    words = [lemmatizer.lemmatize(word) for word in words if word not in stop_words and word.isalpha()]
    return " ".join(words)

# Apply to content column
df['cleaned'] = df['content'].apply(clean_text)
df[['content', 'cleaned']].head()

from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from sklearn.model_selection import train_test_split
from tensorflow.keras.utils import to_categorical
import numpy as np

# Parameters
max_words = 10000  # Max vocab size
max_len = 200      # Max length of each sequence

# Tokenization
tokenizer = Tokenizer(num_words=max_words)
tokenizer.fit_on_texts(df['cleaned'])
sequences = tokenizer.texts_to_sequences(df['cleaned'])

# Padding
X = pad_sequences(sequences, maxlen=max_len)
y = np.array(df['label'])

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, LSTM, Dense, Dropout


max_words = 10000   # Vocabulary size
max_len = 200       # Max length of padded sequences
embedding_dim = 128 # Embedding output dimension

# Build the model
model = Sequential()
model.add(Embedding(input_dim=max_words, output_dim=embedding_dim, input_length=max_len))
model.add(LSTM(128, dropout=0.2, recurrent_dropout=0.2))
model.add(Dense(1, activation='sigmoid'))

model.build(input_shape=(None, max_len))

# Print summary
model.summary()

# Compile the model
model.compile(loss='binary_crossentropy',
              optimizer='adam',
              metrics=['accuracy'])

/usr/local/lib/python3.11/dist-packages/keras/src/layers/core/embedding.py:90: UserWarning: Argument `input_length` is deprecated. Just remove it.
  warnings.warn(

Model: "sequential"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ embedding (Embedding)           │ (None, 200, 128)       │     1,280,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ lstm (LSTM)                     │ (None, 128)            │       131,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense (Dense)                   │ (None, 1)              │           129 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 1,411,713 (5.39 MB)

 Trainable params: 1,411,713 (5.39 MB)

 Non-trainable params: 0 (0.00 B)

history = model.fit(X_train, y_train,
                    epochs=5,
                    batch_size=64,
                    validation_split=0.2)

Epoch 1/5
449/449 ━━━━━━━━━━━━━━━━━━━━ 215s 467ms/step - accuracy: 0.8892 - loss: 0.2390 - val_accuracy: 0.9801 - val_loss: 0.0639
Epoch 2/5
449/449 ━━━━━━━━━━━━━━━━━━━━ 208s 464ms/step - accuracy: 0.9805 - loss: 0.0566 - val_accuracy: 0.9841 - val_loss: 0.0531
Epoch 3/5
449/449 ━━━━━━━━━━━━━━━━━━━━ 209s 465ms/step - accuracy: 0.9888 - loss: 0.0370 - val_accuracy: 0.9847 - val_loss: 0.0525
Epoch 4/5
449/449 ━━━━━━━━━━━━━━━━━━━━ 208s 463ms/step - accuracy: 0.9926 - loss: 0.0258 - val_accuracy: 0.9852 - val_loss: 0.0527
Epoch 5/5
449/449 ━━━━━━━━━━━━━━━━━━━━ 207s 461ms/step - accuracy: 0.9921 - loss: 0.0242 - val_accuracy: 0.9839 - val_loss: 0.0541

import matplotlib.pyplot as plt

# Plot training & validation accuracy
plt.figure(figsize=(12, 5))

plt.subplot(1, 2, 1)
plt.plot(history.history['accuracy'], label='Train Accuracy')
plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
plt.title('Model Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend()

# Plot training & validation loss
plt.subplot(1, 2, 2)
plt.plot(history.history['loss'], label='Train Loss')
plt.plot(history.history['val_loss'], label='Validation Loss')
plt.title('Model Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend()

plt.tight_layout()
plt.show()

def predict_news(text):

    cleaned = clean_text(text)


    seq = tokenizer.texts_to_sequences([cleaned])
    padded = pad_sequences(seq, maxlen=max_len)

    # Predict
    pred = model.predict(padded)[0][0]

    # Interpret result
    label = "REAL" if pred >= 0.5 else "FAKE"
    confidence = pred if pred >= 0.5 else 1 - pred

    print(f"Prediction: {label} ({confidence*100:.2f}% confidence)")

# Example usage
sample_text = """House Republicans float compromise to placate warring factions: Faster Medicaid cuts and a larger SALT deduction
"""
predict_news(sample_text)

1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 427ms/step
Prediction: REAL (97.48% confidence)

model.save('lstmmodel.keras')

from google.colab import files
files.download("lstmmodel.keras")

from tensorflow.keras.models import load_model

# Load the model
loaded_model = load_model("lstmmodel.keras")

# Predict using the loaded model
sample_text = "House Republicans float compromise to placate warring factions: Faster Medicaid cuts and a larger SALT deduction"
predict_news(sample_text)  # Still works — same tokenizer and preprocess

1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 114ms/step
Prediction: REAL (97.48% confidence)

import pickle

# Save tokenizer
with open('tokenizer.pkl', 'wb') as f:
    pickle.dump(tokenizer, f)

# Download tokenizer
files.download("tokenizer.pkl")

# Load tokenizer
with open('tokenizer.pkl', 'rb') as f:
    tokenizer = pickle.load(f)

from google.colab import files
uploaded = files.upload()

Saving FakeNewsDetection.ipynb to FakeNewsDetection (4).ipynb

!jupyter nbconvert --to html FakeNewsDetection (1).ipynb

/bin/bash: -c: line 1: syntax error near unexpected token `('
/bin/bash: -c: line 1: `jupyter nbconvert --to html FakeNewsDetection (1).ipynb'

files.download('FakeNewsDetection (1).html')

Refined Category	Methods & Libraries Used
Management of Data	Pandas, NumPy
Text Cleaning	NLTK, RegEx, WordNet Lemmatizer
Model Development	TensorFlow and Keras
Data Representation	Visualization, Chart.js, Matplotlib and Seaborn
Website Development	Flask, HTML, CSS
Response Storage	SQLite
Web Application	Docker

	title	text	subject	date	label
0	Ben Stein Calls Out 9th Circuit Court: Committ...	21st Century Wire says Ben Stein, reputable pr...	US_News	February 13, 2017	0
1	Trump drops Steve Bannon from National Securit...	WASHINGTON (Reuters) - U.S. President Donald T...	politicsNews	April 5, 2017	1
2	Puerto Rico expects U.S. to lift Jones Act shi...	(Reuters) - Puerto Rico Governor Ricardo Rosse...	politicsNews	September 27, 2017	1
3	OOPS: Trump Just Accidentally Confirmed He Le...	On Monday, Donald Trump once again embarrassed...	News	May 22, 2017	0
4	Donald Trump heads for Scotland to reopen a go...	GLASGOW, Scotland (Reuters) - Most U.S. presid...	politicsNews	June 24, 2016	1

	content	label
0	Ben Stein Calls Out 9th Circuit Court: Committ...	0
1	Trump drops Steve Bannon from National Securit...	1
2	Puerto Rico expects U.S. to lift Jones Act shi...	1
3	OOPS: Trump Just Accidentally Confirmed He Le...	0
4	Donald Trump heads for Scotland to reopen a go...	1

	content	cleaned
0	Ben Stein Calls Out 9th Circuit Court: Committ...	ben stein call circuit court committed coup ét...
1	Trump drops Steve Bannon from National Securit...	trump drop steve bannon national security coun...
2	Puerto Rico expects U.S. to lift Jones Act shi...	puerto rico expects u lift jones act shipping ...
3	OOPS: Trump Just Accidentally Confirmed He Le...	oops trump accidentally confirmed leaked israe...
4	Donald Trump heads for Scotland to reopen a go...	donald trump head scotland reopen golf resort ...

Detecting Fake News using LSTM Deep Learning Model with Real-Time Explanation Capabilitiesn

🧠 Fake News Detection using LSTM - Final Project Report

2. 📍 Testing Methods and Concepts

3. 🗄 Data Sets Used

4. 🗒 Model Structure - LSTM

5. 📈 Model Performance

6. 🌐 Application Features

7. 📋 User Feedback System

8. 🚀 Deployment

9. ✅ Results

10. 🔮 Future Works

Model Building

Setup and Imports¶

Upload Data¶

Load and Merge Datasets¶

Clean & Combine Columns¶

Text Preprocessing¶

Text Cleaning Function¶

Tokenization & Padding¶

Train/Test Split¶

Build LSTM Model¶

Train Model¶

Plot Accuracy and Loss¶

Prediction Function¶

Save Model¶

Testing LSTM Model¶