This project implements an end-to-end algorithm for a recommendation system for music recommendations.
Simple memory-based approaches and a Latent Factor Model (LFM) are used for candidate selection. Gradient Boosting from CatBoost is used for re-ranking among the candidates.
Additionally, the work includes a comparison of different algorithms for recommendations and handling compressed sparse matrices for storing information about user-item interactions.
Examples of different models in action are also provided.
- recommendations-project-RubanovVladislav.ipynb: the project
- music_dataset.csv: data (information about user-item interactions)
- tracks_info.csv: information about music tracks