Collaborative filtering board game recommendations from BoardGameGeek data 🎲🤖.
https://cedricblondeau.github.io/board-game-recs/
board-games-recs
is powered by a collaborative filtering ML model. It uses reviews from users with similar tastes to compute similarity between games.
Model is trained with FastAI which provides great tooling for assembling data. See this notebook for more information.
Recommendations are written to a JSON file and exposed through a React-based UI.
lab/ # Jupyter notebooks
ui/ # React demo app
Model uses this wonderful Kaggle dataset from Jesse van Elteren.
Data comes from the awesome BoardGameGeek XML API. It allows reproducing and displaing the data for strictly non-commercial purposes.
Fun and learning. Zero guarantees included.