Spotify Song Analysis by Statistical Machine Learning

This paper aims to study the use of Artificial Intelligence in the music streaming scenario. Specifically, the Spotify playlist creation process is redesigned by employing statistical machine learning algorithms. Spotify is chosen because it is the undisputed leader in music streaming, founding its success on the listening experience offered to its users. The level of personalization of the Spotify service is high and is made possible precisely because of Artificial Intelligence: by collecting masses of data about users and their listening habits, Spotify is able to understand the single user’s interests and make targeted recommendations. Artificial Intelligence is also being used for another aspect: the creation of playlists, which collect different songs in order to satisfy every need or desire of the user. This work aims precisely at identifying the most relevant playlists by "Mood” through machine learning algorithms from a large sample of Spotify songs.