📰 Fake News Detection App

---

🚀 Live Demo

👉 https://fake-news-detector-ujjawal.streamlit.app

---

📌 Problem Statement

In today’s digital world, misinformation spreads rapidly through social media and online platforms.
It becomes difficult for users to verify whether a news article is genuine or fake.

👉 This project aims to automatically detect fake news using Machine Learning and NLP techniques.

---

💡 Solution

We built an AI-powered web application that:

• Analyzes news text
• Extracts content from URLs
• Classifies news as Real or Fake
• Provides a confidence score
• Shows trending real-world news

---

🎯 Features

• 📝 Text-based fake news detection
• 🌐 URL-based article analysis
• 📊 Confidence score with progress bar
• 🔍 Article preview for better understanding
• 🔥 Live trending news using GNews API
• 🎨 Interactive and modern UI (Streamlit)

---

🧠 Tech Stack

🔹 Programming

• Python

🔹 Machine Learning

• Logistic Regression

🔹 NLP

• TF-IDF Vectorization

🔹 Frontend

• Streamlit

🔹 APIs

• GNews API (for trending news)

🔹 Libraries Used

• scikit-learn
• pandas
• numpy
• newspaper3k
• requests

---

📂 Project Structure

fake_news_project/
│
├── app/
│   └── app.py              # Streamlit application
│
├── data/
│   ├── Fake.csv           # Fake news dataset
│   └── True.csv           # Real news dataset
│
├── model/
│   ├── model.pkl          # Trained ML model
│   └── vectorizer.pkl     # TF-IDF vectorizer
│
├── main.py                # Model training script
├── requirements.txt       # Dependencies
└── README.md              # Documentation

---

⚙️ How It Works (Step-by-Step)

1️⃣ Data Collection

• Used two datasets:
  • Fake News
  • Real News

---

2️⃣ Data Preprocessing

• Converted text to lowercase
• Removed special characters using regex
• Cleaned noise and unwanted symbols

---

3️⃣ Feature Engineering

• Applied TF-IDF Vectorizer
• Converted text into numerical vectors

---

4️⃣ Model Training

• Used Logistic Regression
• Split dataset into training and testing
• Achieved high accuracy (~99%)

---

5️⃣ Prediction Pipeline

User input → Clean text → Transform (TF-IDF) → Model prediction

Output:

• Real / Fake
• Confidence score

---

🌐 URL Processing

• Uses newspaper3k library
• Extracts full article text automatically
• Applies same ML pipeline for prediction

---

🔥 Trending News Feature

• Integrated with GNews API
• Fetches real-time headlines
• Displays:
  • Title
  • Source
  • Link

---

📊 Model Performance

• Accuracy: ~99%
• Precision: High
• Recall: High

👉 Note: High accuracy is due to dataset characteristics.

---

⚠️ Limitations

• Model trained on specific dataset → may not generalize fully
• Cannot verify factual correctness (pattern-based prediction)
• Sensitive to writing style differences

---

🚀 Future Improvements

• Use BERT / Transformers for better understanding
• Add explainability (why prediction is fake/real)
• Improve UI with dashboard design
• Add multilingual support
• Store prediction history

---

▶️ Run Locally

1. Clone Repository

git clone https://github.com/your-username/fake-news-project.git
cd fake-news-project

2. Install Dependencies

pip install -r requirements.txt

3. Run App

streamlit run app/app.py

---

☁️ Deployment

Deployed using Streamlit Community Cloud

---

👨‍💻 Author

Ujjawal Shrivastava
Aspiring Data Scientist

---

⭐ Conclusion

This project demonstrates how Machine Learning + NLP can be applied to solve real-world problems like fake news detection.

It showcases:

• End-to-end ML pipeline
• Real-time API integration
• Interactive UI development
• Cloud deployment

---

Screenshots