Skip to content

Latest commit

 

History

History
64 lines (53 loc) · 2.14 KB

README.md

File metadata and controls

64 lines (53 loc) · 2.14 KB

Fake News Detection System

Overview

This project is a Fake News Detection System that classifies news articles as either Real or Fake using Natural Language Processing (NLP) and Machine Learning. It includes:

  • A Jupyter Notebook for training a Logistic Regression model using TF-IDF feature extraction.
  • A Streamlit Web App for users to input news articles and get real-time predictions.

Features

  • Text Preprocessing: Converts text to lowercase, removes special characters, stems words, and eliminates stopwords.
  • TF-IDF Vectorization: Extracts the top 5000 most important words.
  • Machine Learning Model: Logistic Regression classifier for accurate predictions.
  • Model Deployment: A Streamlit web app to check if a news article is real or fake.

Installation

  1. Clone the repository: 
    git clone https://github.com/sakuna47/Spam_Detection
  2. Install dependencies: 
    pip install -r requirements.txt
  3. Download NLTK stopwords: 
    import nltk
nltk.download('stopwords')
  4. Run the Streamlit app: 
    streamlit run app.py

Usage

  1. Train the model using the Jupyter Notebook (fake_news_detection.ipynb).
  2. Save the trained model and vectorizer.
  3. Run app.py and enter a news article to check its authenticity.

Dataset

  • Real News: Sourced from legitimate news websites.
  • Fake News: Collected from unreliable sources.
  • Both datasets are labeled (0 = Real, 1 = Fake).

Technologies Used

  • Python
  • Pandas, NumPy, Scikit-Learn (for data processing & ML)
  • NLTK (for text preprocessing)
  • Matplotlib, Seaborn (for data visualization)
  • Streamlit (for web app deployment)

Results

  • Model Accuracy: ~95%
  • Confusion Matrix & Classification Report provided in the notebook.

Future Improvements

  • Improve model with deep learning (LSTM, BERT).
  • Add multilingual support for news detection.
  • Deploy as a cloud-based API for broader use.

Contributors

License

This project is open-source under the MIT License.