Automated Mood Based Music Playlist Generation By Clustering The Audio Features

The increase of receiving attention to music recommendation and playlist generation in today's music industry is undeniable. One of the main goals is to generate personalized playlists automatically for each user. Beyond that, an appropriate switching among these playlists to play the tracks based on the current mood of the user would certainly lead to the development of more advanced and personalized music player apps. In this paper, a data scientific approach is provided to model the music moods which are created by clustering the tracks extracted from users' listening. Each Cluster consists of music tracks with similar audio features existing in the user's listening history. Knowing which music track is currently being listened by users, their mood would be specified by determining the cluster of that music. It is presumed that playing the other music tracks contained in the same cluster as the next tracks will enhance their satisfaction. A suggestion for making the results visually interpretable which could help the corresponding music players with GUI design is provided as well. Experimental results of a case study from real datasets collected from Users' listening history on Last.fm benefiting from Spotify API clarifies the framework along with supporting the mentioned presumption.