![\"Open](\"https://colab.research.google.com/assets/colab-badge.svg\"){kind=icon}
"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "UaBQOwyL-1s6"
},
"source": [
"# Fake News Detection Using Machine Learning and Deep Learning Models\n",
"\n",
"This notebook aims to classify fake news from real news."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"id": "J0e8axCk-1s-"
},
"outputs": [],
"source": [
"# Importing necessary libraries\n",
"import pandas as pd\n",
"import numpy as np\n",
"\n",
"# Data Visualization\n",
"import matplotlib.pyplot as plt\n",
"import seaborn as sns\n",
"plt.style.use('ggplot')\n",
"\n",
"# Machine Learning\n",
"from sklearn.metrics import accuracy_score, f1_score, classification_report\n",
"from sklearn.model_selection import train_test_split\n",
"from sklearn.utils import resample\n",
"from sklearn.feature_extraction.text import TfidfVectorizer # Import for transforming text into numerical features using TF-IDF\n",
"from sklearn.linear_model import LogisticRegression\n",
"from sklearn.ensemble import RandomForestClassifier\n",
"\n",
"# Deep Learning\n",
"import tensorflow as tf # Import TensorFlow library for deep learning tasks\n",
"import keras # Import Keras library for building neural networks\n",
"from tensorflow.keras.preprocessing.sequence import pad_sequences # Import for padding sequences\n",
"from keras.models import Model, Sequential # Import for defining neural network models\n",
"from keras.layers import Input, Dense, Activation, Bidirectional, LSTM, Dropout, Embedding # Import for different layers in neural networks\n",
"from keras.preprocessing.text import Tokenizer # Import for tokenizing text\n",
"from keras.preprocessing import sequence # Import for processing sequences\n",
"from keras.callbacks import EarlyStopping # Import for early stopping during model training\n",
"\n",
"# Configuring TensorFlow logging\n",
"tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR) # Set TensorFlow logging verbosity to error level\n",
"\n",
"# Text processing\n",
"from wordcloud import WordCloud # Import for generating word clouds\n",
"\n",
"# Mounting Google Drive\n",
"from google.colab import drive # Import for accessing Google Drive\n",
"\n",
"# Unzipping files\n",
"import zipfile # Import for extracting zip files\n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Z3rDHoCj-1tB"
},
"source": [
"Read the data"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "qiMxcfZE-1tC",
"outputId": "ac5873fa-0d5a-4b79-b28d-04d3146dac57"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"Mounted at /content/drive\n"
]
}
],
"source": [
"# Mount your Google Drive\n",
"drive.mount('/content/drive')\n",
"\n",
"# Get the file path from Google Drive\n",
"file_path = '/content/drive/MyDrive/fake news/archive (2).zip'\n",
"\n",
"# Unzip the file\n",
"with zipfile.ZipFile(file_path, 'r') as zip_ref:\n",
" # Find the CSV files in the zip folder\n",
" fake_path = zip_ref.extract('Fake.csv', '/content/')\n",
" real_path = zip_ref.extract('True.csv', '/content/')\n",
"\n",
"# Read the csv file from the url\n",
"fake_df = pd.read_csv(fake_path)\n",
"real_df = pd.read_csv(real_path)\n",
"\n",
"# A way to delete rows with empty or null values\n",
"fake_df = fake_df[~fake_df.isna().any(axis=1)]\n",
"real_df = real_df[~real_df.isna().any(axis=1)]"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "LXTkxHhL-1tE"
},
"source": [
"Checking for null values"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "x7dA9dXK-1tE",
"outputId": "0ebe326b-ec44-4967-b2ae-82d0e03d868b"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"title 0\n",
"text 0\n",
"subject 0\n",
"date 0\n",
"dtype: int64"
]
},
"metadata": {},
"execution_count": 3
}
],
"source": [
"fake_df.isnull().sum()"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "YRdyEUyd-1tF",
"outputId": "f6994243-9132-44e4-b1d1-08ccaff6e4cd"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"title 0\n",
"text 0\n",
"subject 0\n",
"date 0\n",
"dtype: int64"
]
},
"metadata": {},
"execution_count": 4
}
],
"source": [
"real_df.isnull().sum()"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "or_wLYlt-1tG"
},
"source": [
"Checking for unique values for subject. We want both data frames to have a similar distribution."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "4VCzCS3S-1tG",
"outputId": "33d3cfa8-e70c-48ae-8c8b-dc316eb33570"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"array(['News', 'politics', 'Government News', 'left-news', 'US_News',\n",
" 'Middle-east'], dtype=object)"
]
},
"metadata": {},
"execution_count": 5
}
],
"source": [
"fake_df.subject.unique()"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "ah3MTyim-1tH",
"outputId": "b1a06e2f-e0e2-494b-a54e-a60252f621a7"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"array(['politicsNews', 'worldnews'], dtype=object)"
]
},
"metadata": {},
"execution_count": 6
}
],
"source": [
"real_df.subject.unique()"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "gPOdec7r-1tI"
},
"source": [
"Drop the date from the dataset, I don't think there is a strong correlation between date and validity of the news. As we see above, subjects are not distributed evenly. We do not want that to influence the accuracy of our classifier. Therefore, we need to drop that as well."
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"id": "qTTMeqws-1tI"
},
"outputs": [],
"source": [
"fake_df.drop(['date', 'subject'], axis=1, inplace=True)\n",
"real_df.drop(['date', 'subject'], axis=1, inplace=True)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "HlFnS5u9-1tJ"
},
"source": [
"0 for fake news, and 1 for real news"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"id": "URgIOOZM-1tJ"
},
"outputs": [],
"source": [
"fake_df['class'] = 0\n",
"real_df['class'] = 1"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "M9Kce-4HFdWI"
},
"source": [
"## EDA"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "D5UKQo1f-1tJ"
},
"source": [
"Check out the distribution of fake news compare to real news"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 990
},
"id": "x1FnUH17-1tK",
"outputId": "df59c638-a9b2-47ca-b035-e852f0e96f94"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"Text(0, 0.5, 'Proportion of News Articles')"
]
},
"metadata": {},
"execution_count": 9
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"| \n", " | title | \n", "text | \n", "class | \n", "
|---|---|---|---|
| 0 | \n", "Donald Trump Sends Out Embarrassing New Year’... | \n", "Donald Trump just couldn t wish all Americans ... | \n", "0 | \n", "
| 1 | \n", "Drunk Bragging Trump Staffer Started Russian ... | \n", "House Intelligence Committee Chairman Devin Nu... | \n", "0 | \n", "
| 2 | \n", "Sheriff David Clarke Becomes An Internet Joke... | \n", "On Friday, it was revealed that former Milwauk... | \n", "0 | \n", "
| 3 | \n", "Trump Is So Obsessed He Even Has Obama’s Name... | \n", "On Christmas day, Donald Trump announced that ... | \n", "0 | \n", "
| 4 | \n", "Pope Francis Just Called Out Donald Trump Dur... | \n", "Pope Francis used his annual Christmas Day mes... | \n", "0 | \n", "
| ... | \n", "... | \n", "... | \n", "... | \n", "
| 44893 | \n", "'Fully committed' NATO backs new U.S. approach... | \n", "BRUSSELS (Reuters) - NATO allies on Tuesday we... | \n", "1 | \n", "
| 44894 | \n", "LexisNexis withdrew two products from Chinese ... | \n", "LONDON (Reuters) - LexisNexis, a provider of l... | \n", "1 | \n", "
| 44895 | \n", "Minsk cultural hub becomes haven from authorities | \n", "MINSK (Reuters) - In the shadow of disused Sov... | \n", "1 | \n", "
| 44896 | \n", "Vatican upbeat on possibility of Pope Francis ... | \n", "MOSCOW (Reuters) - Vatican Secretary of State ... | \n", "1 | \n", "
| 44897 | \n", "Indonesia to buy $1.14 billion worth of Russia... | \n", "JAKARTA (Reuters) - Indonesia will buy 11 Sukh... | \n", "1 | \n", "
44898 rows × 3 columns
\n", "LogisticRegression()In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
LogisticRegression()
RandomForestClassifier(random_state=0)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
RandomForestClassifier(random_state=0)