Music recommender system from Spotity playlists dataset
Made for the initiative project-of-the-week from DataTalks.club https://github.com/DataTalksClub/project-of-the-week/blob/main/2022-10-19-recommenders-1.md
Spotify Playlists. 1.2GB of tabular data for music recommendation https://www.kaggle.com/datasets/andrewmvd/spotify-playlists
- Select dataset
- Create repo in git
- Clone repo and setup environment (pipenv, install libraries)
- Download dataset
- EDA in jupyter notebook
- Analyse sparcity
- Set a baseline
- Set model evaluation metrics
- Analyse solutions
- Implement matrix factorization model with LightFM
- [ ]
- In collaborative filtering, consider weight samples in order to avoid frequent queries or frequent items dominate the objective function
- Consider cold start problem in collaborative filtering: https://developers.google.com/machine-learning/recommendation/collaborative/summary?hl=en
Update pip on Windows: python -m pip install --upgrade pip
Update pipenv before creating the virtual env: pip install --upgrade pipenv
Create basic virtual env: pipenv install pandas numpy --dev ipython