Book Recommender

Book recommendation system using Deep Learning

Made by:

1. Oksana Konovalova, [email protected]
2. Adelina Kildeeva, [email protected]

Project topic

Recommendation systems can be found in almost every area of life, from shopping on marketplaces to choosing a TV show for the weekend. We wanted to study the topic of recommendations in more detail, understand the work of various methods and create our own recommendation system for books. The goal of our project is to study and develop a hybrid book recommendation system.

Dataset

To train and evaluate our systems, we used the open goodbooks-10k dataset, consisting of 10000 books, 53424 users and 6 million ratings. The description and analysis of the data can be found in the 1.0-initial_data_exploration.ipynb

Demo

Full demo video with good quality

Models Comparison

	Simple CF	Content-based system	User-based CF using DL	Hybrid model	Fine-tuned hybrid model
RMSE	3.53569	0.95774	0.98924	0.89941	0.87325
RMSE for high-rated data	---	0.89935	0.81974	0.76834	0.74349

How to use

First option: use the telegram bot

https://t.me/ultimate_book_recommender_bot

Second option: run the bot locally

1. Clone the repository

2. Get a telegram token and put it in the 'boot_key.key' file.

3. Install the required packages

pip install -r requirements.txt

4. Run the bot

python bot.py

Learn more

Learn more about the project here