https://github.com/philiptitus/collaborative-book-recommender
Made use of the content-based filtering algorithm to make a book recommender model
https://github.com/philiptitus/collaborative-book-recommender
content-based-filtering content-based-recommendation recommender-system tensorflow
Last synced: about 2 months ago
JSON representation
Made use of the content-based filtering algorithm to make a book recommender model
- Host: GitHub
- URL: https://github.com/philiptitus/collaborative-book-recommender
- Owner: philiptitus
- License: mit
- Created: 2025-04-02T10:20:31.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-04-02T10:24:09.000Z (2 months ago)
- Last Synced: 2025-04-02T11:28:35.516Z (2 months ago)
- Topics: content-based-filtering, content-based-recommendation, recommender-system, tensorflow
- Language: Jupyter Notebook
- Homepage:
- Size: 10.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Book Recommendation System
A deep learning-based collaborative filtering recommendation system for books using the Book-Crossing dataset. This system predicts user ratings for books based on both user and book features.
## Overview
This project implements a neural network-based recommendation system that:
- Uses collaborative filtering to predict book ratings
- Incorporates both user and book features
- Provides personalized book recommendations
- Achieves good prediction accuracy on the test set
## Features
- User embedding layer for learning user preferences
- Item embedding layer for learning book characteristics
- Dot product layer for rating prediction
- Regularization to prevent overfitting
- Data preprocessing and feature engineering
- Model evaluation and recommendation generation
## Dataset
The model uses the Book-Crossing dataset from Kaggle, which contains:
- Book information (ISBN, title, author, year of publication, publisher)
- User information (user ID, location, age)
- Rating information (user ratings for books)
## Setup
1. Clone the repository:
```bash
git clone
cd book-recommendation-system
```
2. Create a virtual environment (recommended):
```bash
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
```
3. Install dependencies:
```bash
pip install -r requirements.txt
```
4. Place the dataset files in the `data` directory:
- BX_Books.csv
- BX-Users.csv
- BX-Book-Ratings.csv
## Usage
1. Run the Jupyter notebook:
```bash
jupyter notebook model.ipynb
```
2. Execute all cells in sequence to:
- Load and preprocess the data
- Train the model
- Generate recommendations
## Model Architecture
The recommendation system uses a neural network with:
- User embedding layer (256 → 128 → 32 neurons)
- Item embedding layer (256 → 128 → 32 neurons)
- Dot product layer for rating prediction
- L2 normalization for embeddings
## Performance
The model achieves:
- Training loss: ~0.56
- Test loss: ~0.57
- Efficient training on 66,003 samples
- Good generalization on test set
## Author
Philip Titus
## Connect with Me
- 🌐 [Personal Website](https://mrphilip.pythonanywhere.com/)
- 🛍️ [Shop](https://pmart-pi.vercel.app/)
- 👥 [LinkedIn](https://linkedin.com/in/philiptitus)
## License
Copyright © 2025 Philip Titus. All rights reserved.