/music_recommendation_system

This project proposes a recommendation system with top 10 songs for a user based on the likelihood of listening to those songs in the audio content provider Spotify.

Primary LanguageJupyter Notebook

music_recommendation_system

This project proposes a recommendation system with top 10 songs for a user based on the likelihood of listening to those songs in the audio content provider Spotify. Based on training the available dataset with various recommended system models, we infer that the user-user similarity based recommended system provides the highest F1 score and is most appropriate for the data. The model invoked in the user-user similarity recommended system is a basic k-nearest neighbor model with optimal hyperparameters being : maximum number of nearest neighbors = 30, minimum number of neighbors = 9 and the Pearson baseline method to quantify the similarity between 2 users. We further adapat this model to also take into account the popularity of song among various users.