Hybrid Book Recommender System

This is a Hybrid Recommendation System deployed on Hugging Face Spaces using Gradio.
It combines Collaborative Filtering (SVD++) and Content-Based Filtering to recommend books to users, even if they are new to the system (cold-start support).

🔗 Live Demo: Check it out on Hugging Face Spaces

Features

- Hybrid Model: Combines collaborative filtering and content-based filtering for more accurate recommendations.  
- Cold Start Ready: Recommends top-rated books to new users with no prior ratings.  
- Book Covers: Shows book covers alongside titles for a more visual experience.  
- Gradio: Backend runs on FastAPI, frontend built with Gradio for an interactive experience.  
- Model loaded from Kaggle Models using `kagglehub`.  

How It Works

1. For registered users:
   - Predicts ratings for books the user hasn’t rated yet.
   - Shows the top N books sorted by predicted rating.

2. For new users:
   - Recommends globally top-rated books as a fallback (cold start).  

3. Visual Output:
   - Book title, predicted rating, and cover image in a neat table.

Project Structure

- ├── app.py # Main Gradio app
- ├── requirements.txt # Dependencies
- ├── README.md
- ├── books-recomendation-system-0.ipynb => Model Development Notebook

Setup Locally

Clone the repo and run:

pip install -r requirements.txt
python app.py

Dependencies

- gradio
- pandas
- numpy
- scikit-learn
- kagglehub
- surprise (for SVD++ model)
- fastapi
- uvicorn

Model Details

- Collaborative Filtering: SVD++ trained on user-item ratings.
- Content-Based Filtering: Uses book metadata (title, genres, etc.).
- Hybrid approach balances both for better personalization.

Deployment

- This app is deployed on Hugging Face Spaces using Gradio UI and loads the model directly from Kaggle Models via kagglehub.