MusicMagal is a group recommendation system that recommends n music tracks to a group of m users considering all of the m users preferences into account. To achieve this we've based our machine learning and deep learning models in Last.Fm data. After computing and when the resulting playlist is output, we create a real playlist using Spotify API's python wrapper: Spotipy.
A typical version of the program's flow is presented in musicmagal_flow.ipynb.
To see our evaluation metrics you can check musicmagal_evaluation.ipynb.
Our database exploration is presented in db_exploration.ipynb.
You can read the article going over our model bit by bit in Hacker Noon.
[1] Mezei, Zsolt, and Carsten Eickhoff. "Evaluating Music Recommender Systems for Groups." arXiv preprint arXiv:1707.09790 (2017).
[2] Yoshii, Kazuyoshi, et al. "Hybrid Collaborative and Content-based Music Recommendation Using Probabilistic Model with Latent User Preferences." ISMIR. Vol. 6. 2006.
[3] Parra, Denis, et al. "Implicit feedback recommendation via implicit-to-explicit ordinal logistic regression mapping." Proceedings of the CARS-2011 (2011).
[4] Hu, Yifan, Yehuda Koren, and Chris Volinsky. "Collaborative filtering for implicit feedback datasets." Data Mining, 2008. ICDM'08. Eighth IEEE International Conference on. Ieee, 2008.
[5] Barkan, Oren, and Noam Koenigstein. "Item2vec: neural item embedding for collaborative filtering." Machine Learning for Signal Processing (MLSP), 2016 IEEE 26th International Workshop on. IEEE, 2016.
[6] Leskovec, Jure, Anand Rajaraman, and Jeffrey David Ullman. 'Mining of massive datasets.' Cambridge university press, 2014.