An approach for music recommendation using content-based analysis and collaborative filtering

In this paper, we propose a new approach for music recommendation using a content-based analysis and collaborative filtering. The proposed method uses the playlists of other users as collaborative filtering approaches do, but it partitions each playlist into similar music groups. Then, it finds the music groups which are very similar to each of music groups of the target user. The method recommends pieces of music which the target user has not yet listened in the selected groups. For partitioning, we use a content-based analysis approach which is based on the MFCCs and the HMMs to calculate the music similarity. Through our method, we can recommend proper pieces of music which may satisfy the target user. The proposed method shows better performance when compared to contentbased analysis and collaborative filtering approaches.