A music recommendation system based on collaborative filtering and SVD

With the development of the Internet and the advent of music-streaming platforms, a large amount of music data available for selection is now greater than ever on the Internet. In addition to searching expected music objects for users, it becomes necessary to develop a recommendation service. A music recommendation system (MRS) relieves users from sorting through the various options by automatically recommending music based on their historical behaviors like the play count of each song. Recommender systems have aroused a lot of awareness in the past decade. Although algorithms including content-based, collaborative, singular value decomposition, and other techniques are used in the recommendation system, there does not exist any perfect recommendation system that can give completely precise feedback on what users actually want. To figure out which algorithm does a better job, the paper proposes a music recommendation system based on two algorithms, item-based collaborative filtering, and singular value decomposition, that are used in the music recommendation system and compares the two methods to find out which one can make a more precise recommendation. Item similarity between the songs listened by the user and other users is used to predict which songs are preferred by the user. In this paper, D-Recall is regarded as an evaluation indicator between the two algorithms. And the performance of SVD is better than item-based collaborative filtering on the recommendation.