Fake_News_Detector

📰 Fake News Detector — Streamlit App

A lightweight NLP-powered web app that classifies news articles as Real or Fake using an LSTM-based deep learning model. The dataset is hosted on Hugging Face, and the app is fully deployable on Streamlit Cloud.

---

🚀 Features

• End-to-end text preprocessing (cleaning, tokenization, lemmatization)
• LSTM neural network trained on a custom fake news dataset
• Hugging Face dataset loading via direct raw URLs
• Interactive Streamlit UI for predictions
• Real-time confidence scoring and visualization

---

📂 Project Structure

.
├── app.py
├── requirements.txt
├── README.md

---

🛠 Tech Stack

• Frontend: Streamlit
• NLP: NLTK
• Data Processing: Pandas, NumPy
• Modeling: TensorFlow, Keras, Scikit-learn
• Visualization: Matplotlib, Seaborn
• Dataset Hosting: Hugging Face Datasets

---

📦 Dataset

Dataset used for training: https://huggingface.co/datasets/AyushiAyushi3017/fake-news-detector-dataset

Contains two CSV files: Fake.csv and True.csv.

---

🧠 Model Overview

• Tokenizer + padded sequences
• Embedding layer (100 dimensions)
• LSTM layer (150 units)
• Dropout regularization
• Global Max Pooling
• Dense layers with ReLU and Softmax
• Trained for 15 epochs with 80/20 train-test split

---

▶️ Running Locally

1. Clone the repository

git clone https://github.com/<your-username>/<repo>.git
cd <repo>

2. Install dependencies

pip install -r requirements.txt

3. Run the Streamlit app

streamlit run app.py

---

🌐 Deployment

Fully compatible with Streamlit Cloud. Push your repo → connect Streamlit → deploy.

---

📌 Future Enhancements

• Switch to transformer-based models (BERT/DistilBERT)
• Add URL-based scraping and prediction
• Add a REST API endpoint
• Improve UI design and styling

---

🤝 Acknowledgements

Built using open-source tools and a custom Hugging Face dataset.