📰 Fake News Detection using Machine Learning & NLP

A Machine Learning project that detects real vs fake news articles using Natural Language Processing (NLP) and Flask web deployment. The model achieves ~99.37% accuracy using the PassiveAggressiveClassifier and TF-IDF vectorization.

---

🚀 Project Overview

The project aims to combat misinformation by building a text classification system that can identify whether a given news article or headline is real or fake. It follows a complete ML pipeline — from data preprocessing to model training, evaluation, and Flask deployment.

---

🧠 Key Features

• 🧹 Text Preprocessing with NLTK (stopword removal, stemming, normalization)
• 🧾 TF-IDF Vectorization for feature extraction
• ⚡ PassiveAggressiveClassifier for efficient text classification
• 📊 Evaluation using Accuracy, Confusion Matrix, ROC Curve, and Precision-Recall Curve
• 🌐 Flask Web App for real-time predictions
• 💾 Model Persistence using joblib

---

📂 Project Structure

Fake-News-Detection/
│
├── dataSets/
│   ├── True.csv
│   ├── Fake.csv
│   └── preprocessed.csv
│
├── model/
│   ├── fake_news_model.pkl
│   └── tfidf_vectorizer.pkl
│
├── src/
│   ├── preprocess.py
│   ├── train_model.py
│   └── evaluate.py
│
├── templates/
│   └── index.html
│
├── static/
│   └── style.css
│
├── app.py
├── requirements.txt
└── README.md

---

⚙️ Installation & Setup

1️⃣ Clone the repository

git clone https://github.com/Murali-KrishnaM/Fake_News_Detector
cd Fake-News-Detection

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Preprocess the dataset

python src/preprocess.py

4️⃣ Train the model

python src/train_model.py

5️⃣ Evaluate performance

python src/evaluate.py

6️⃣ Run the Flask app

python app.py

Then visit http://127.0.0.1:5000/ in your browser.

---

🧩 Tools & Libraries Used

Category	Libraries
Data Handling	pandas, numpy
NLP	nltk
Machine Learning	scikit-learn
Visualization	matplotlib, seaborn
Deployment	flask
Model Storage	joblib

---

📈 Model Performance

• Accuracy: ~99.37%
• Algorithm: PassiveAggressiveClassifier
• Feature Extraction: TF-IDF Vectorization
• Evaluation Metrics: Accuracy, F1-Score, ROC, Confusion Matrix

---

💻 Flask Web App Preview

• 🟢 Real News → Model predicts genuine articles
• 🔴 Fake News → Model flags potentially false content

The frontend is built using simple HTML/CSS with a clean and responsive design.

---

🧪 Future Improvements

• Integrate Deep Learning models (e.g., LSTM, BERT)
• Deploy to Render, Railway, or AWS for public access
• Expand dataset for better generalization
• Add confidence score visualization on predictions

---

👨‍💻 Author

Murali Krishna
B.Tech Artificial Intelligence & Data Science
Rajalakshmi Engineering College, Chennai

ML Enthusiast | Flask Developer

🔗 LinkedIn • 💻 GitHub

---

⭐ If you find this project useful, don't forget to star the repo!