This repository contains the code for the Recommender Systems challenge (Politecnico di Milano, 2023-2024), made on Kaggle.
The application domain is book recommendation, and the goal of the competition is to discover which items a user will interact with.
The datasets contains interactions of users with books: if the user attributed to the book a rating of at least 4, the interaction is present in the dataset with value 1.
The final recommender used in the challenge is an hybrid of:
- SLIM Elastic Net
- RP3beta
obtained merging the two similarity matrices using a weighted sum.
All the models have been trained using both Kaggle and Colab notebooks.
The evaluation metric used for the competition is MAP@10.
- Public leaderboard score: 0.14042 (18th / 63)
- Private leaderboard score: 0.13984 (20th / 63)
The code in this repository was built upon the course repository, which provides the implementation of recommenders and utility code.