🔍 Fake News Detection System

An advanced machine learning application that uses Natural Language Processing (NLP) techniques to detect fake news articles. The system provides both a web interface and a desktop GUI for easy testing and deployment.

✨ Features

• Advanced NLP Analysis: Uses TF-IDF vectorization and SGD Classifier (Linear SVM)
• Model Explainability: Highlights the top contributing keywords and linguistic patterns that influence the "Fake" or "Real" verdict.
• Interactive Dashboard: Modern web interface with confidence visualization and real-time progress bars.
• Fact-Checking Integration: Direct links to Google News and Snopes for cross-referencing news headlines instantly.
• Scan History: Persistent tracking of your recent news analyses within the browser (using local storage) or within the GUI application.
• Improved Confidence Scoring: Uses a sigmoid-based probability mapping for more accurate certainty levels.
• Secure Model Serialization: Custom JSON-based serialization for the model and vectorizer to prevent security risks associated with legacy formats.
• Dual Interface: Fully functional Flask-based Web App and a Tkinter-based Desktop GUI with dashboard layouts.

🏗️ Model Architecture & Workflow

graph TD
    A[Input News Article] --> B[Text Preprocessing]
    B --> C[TF-IDF Vectorization]
    C --> D[SGD Classifier / Linear SVM]
    D --> E{Prediction}
    E -->|0| F[FAKE News]
    E -->|1| G[REAL News]
    
    subgraph "Preprocessing Steps"
    B1[Lowercase Conversion]
    B2[Remove URLs & Mentions]
    B3[Remove Special Characters]
    B4[Eliminate Stopwords]
    B5[Porter Stemming]
    B --> B1 --> B2 --> B3 --> B4 --> B5
    end
    
    subgraph "Security Layer"
    S1[Secure JSON Serialization]
    D -.-> S1
    C -.-> S1
    end

Loading

🚀 Quick Start

Prerequisites

• Python 3.8+
• pip (Python package installer)

Installation

1. Clone the repository:

git clone https://github.com/Ghulam-Mustafa-Keerio/Fake-News_Detection_System.git
cd Fake-News_Detection_System

2. Install required dependencies:

pip install -r requirements.txt

3. Train the model:

python train_model.py

This will generate model.json and vectorizer.json files using a Secure Serialization Format.

💻 Usage

Web Application

1. Start the Flask server:

python app.py

2. Open your browser and navigate to http://localhost:5000

Desktop GUI (Tkinter)

Run the standalone desktop application:

python gui.py

🧠 How It Works

1. Data Processing

The system handles large datasets including fake.csv and true.csv, though for demonstration purposes, train_model.py includes a sample data generator.

2. Text Preprocessing

• Cleaning: Removes noise such as HTML tags, URLs, and special characters.
• Tokenization & Normalization: Splits text into words and applies lowercase conversion.
• Stemming: Reduces words to their base form (e.g., "running" becomes "run").

3. Feature Extraction

• TF-IDF (Term Frequency-Inverse Document Frequency): A statistical measure used to evaluate how important a word is to a document in a collection.
• N-Grams: Uses both unigrams and bigrams (1, 2) for better contextual understanding.

4. Classification Engine

• SGD Classifier: Implements a Linear SVM with hinge loss, optimized for sparse text data.
• Explainability Layer: The model now identifies and extracts the top 5 words that contributed most to a "Fake" or "Real" verdict.
• Robust Confidence Scoring: Implements a sigmoid-based probability mapping ($1 / (1 + e^{-k \cdot distance})$) for more realistic certainty levels.
• Secure Storage: Models are saved as human-readable JSON files, eliminating the pickle security vulnerabilities.

📊 Model Performance & Evaluation

The following results were obtained from the current model training session:

📊 Training Highlights

• Model Accuracy: 71.43%
• Samples: 70 total (35 Fake, 35 Real)
• Train/Test Split: 80/20 ratio (14 samples for evaluation)

📋 Classification Report

Class	Precision	Recall	F1-Score	Support
Fake	0.88	0.70	0.78	10
Real	0.50	0.75	0.60	4
Overall Accuracy			0.71	14

🔍 Live Predictions Analysis

News Text Sample	Prediction	Confidence	Contributing Keywords
"Scientists make breakthrough in cancer research at local university"	REAL	93.63%	local, scientists, research
"Aliens control the government and hide among us confirmed"	FAKE	98.46%	control, government
"Local community celebrates new park opening with festival"	REAL	99.96%	local, new, community

�📁 Project Structure

Fake-News_Detection_System/
├── train_model.py          # Model training & secure serialization
├── app.py                  # Flask web server
├── gui.py                  # Tkinter UI application
├── requirements.txt        # Package dependencies
├── fake.csv & true.csv     # Training datasets
├── templates/             # HTML for web interface
├── model.json             # Securely saved model weights
└── vectorizer.json        # Securely saved TF-IDF vocabulary 

🛠️ Technologies Used

• Python: Core programming language
• Flask: Web framework for REST API
• Scikit-learn: Machine learning library
• NLTK: Natural language processing toolkit
• Pandas & NumPy: Data manipulation
• Tkinter: Desktop GUI framework
• JSON: Secure model serialization (replaced Pickle/Joblib)

📊 Sample Usage

Python API Example

from train_model import FakeNewsDetector

# Initialize detector
detector = FakeNewsDetector()
detector.load_model()

# Make prediction
text = "Your news article text here..."
prediction, confidence, keywords = detector.predict(text)

label = "REAL" if prediction == 1 else "FAKE"
print(f"Result: {label} (Confidence: {confidence:.2f}%)")

# View keyword impact
for kw in keywords:
    print(f"- {kw['word']}: {kw['impact']} (Impact Score: {kw['score']})")

🔐 Security Benefits

• This system is an assistive tool, not a definitive fact-checker
• Always verify important news from multiple reliable sources
• Model accuracy depends on training data quality and diversity
• The system works best with English language news articles

🤝 Contributing

Contributions are welcome! Please feel free to submit issues and pull requests.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Built with open-source machine learning libraries
• Inspired by the need to combat misinformation
• Thanks to the NLP and ML communities for their tools and resources

📧 Contact

For questions or suggestions, please open an issue on GitHub.

---

Note: This is an educational project demonstrating NLP and machine learning applications. For production use, consider using larger, more diverse training datasets and additional validation mechanisms.