Fake News Headline Classifier

A command-line machine learning tool that classifies news headlines as Real or Fake using Logistic Regression and TF-IDF vectorization.

Built as part of the Fundamentals of AI and ML course project.

---

What It Does

You type in a news headline. The model tells you whether it looks like real news or fake news — along with a confidence score.

Enter a headline: Scientists discover cure hidden by Big Pharma

Result     : FAKE
Confidence : 87.3%
Fake prob  : 87.3%  |  Real prob: 12.7%

---

How It Works

1. A dataset of real and fake headlines is vectorized using TF-IDF (Term Frequency-Inverse Document Frequency), which converts text into numerical features based on word importance.
2. A Logistic Regression model is trained on these features to learn patterns that distinguish fake headlines from real ones.
3. When you enter a new headline, the model predicts its label and outputs a probability.

---

Tech Stack

• Python 3
• scikit-learn (Logistic Regression, TF-IDF, train/test split)
• NumPy

---

Setup

1. Clone the repository

git clone https://github.com/prinxeeee/fake-news-classifier.git
cd fake-news-classifier

2. Install dependencies

pip install scikit-learn numpy

3. Run the classifier

python fake_news_detector.py

---

How to Use

When you run the script, it first trains the model and shows accuracy, then prompts you:

==================================================
       FAKE NEWS HEADLINE CLASSIFIER
==================================================

Model trained on 30 headlines
Test Accuracy: 83.3%

==================================================
  Try it yourself! Enter a headline below.
  Type 'quit' to exit.
==================================================

Enter a headline:

Type any headline and press Enter. Type quit to exit.

Example Inputs to Try

SHOCKING: Government puts mind control chips in vaccines

Researchers publish new study on climate change effects

EXPOSED: Secret cure for cancer hidden by Big Pharma

City council approves plan to expand public transport

How It Works

Feature Extraction

• Input headlines are lowercased and stop words are removed
• TfidfVectorizer converts each headline into a numerical feature vector
• Words that are rare but distinctive get higher weight

Classification

• LogisticRegression learns the boundary between fake and real patterns
• Trained on an 80/20 train-test split
• Outputs both a label and a probability score

---

Project Structure

fake-news-classifier/
│
├── fake_news_detector.py   # Main script
└── README.md               # This file

---

Limitations

• Trained on a small dataset of 30 headlines — a real-world version would use thousands of examples
• Works best with English headlines
• Neutral-toned fake headlines may not be caught reliably

Real-World Applications

• Browser extensions that flag suspicious headlines
• Social media misinformation filters
• Journalism tools for quick credibility checks
• Educational NLP demos

Author

Name: Prince Mahar

Registration No: 25BCE10429

Branch: B.Tech CSE (Core)

University: VIT Bhopal University

Course: CSA2001 - Fundamentals of AI and ML

License

This project is submitted as part of the BYOP (Bring Your Own Project) assignment for academic purposes.