Music recommendation based on embedding model with user preference and context
Lei Jin, Dongfeng Yuan, Haixia Zhang · 2017
With the development of the Internet and mobile network technology, it is becoming more and more important to recommend what users want from numerous music. In this paper, we present a kind of embedding model based on user preference and context information, to recommend sequential music playlists to users. We regard songs, users, and contextual labels as points embedded in a Euclidean space through a Markov chain model, and the distances among the points in the embedding model reflect the relationship among songs, users, and contextual labels. Given a song, a playlist with the highest probability will be generated according to the embedding model. In the end, an experiment based on real data demonstrated that this model can improve music recommendation performance.