Research Of Social Network Based Personalized Music Recommendation Algorithm

With the development of Internet, the number of digital tracks is growing exponentially. This makes users to find and search tracks very difficultly. For this reason, scholars pay more and more attention to music retrieval. Traditional music retrieval is based on metadata. It requires users to remember some information to retrieve target tracks. This method is no longer suitable for the fast pace of users’life. Most cases, users only need background music. With the emergence of music recommendation system, users enable gain music that conforms to their interest quickly and continuously. However, how to provide different users different music list that they like become the major issue to be addressed.Web2.0and social media allow users’free behavior. Users can define tags to music rely on their own understanding, and different tags can interprets users’feeling of music from different angles. Therefore, we proposed personalized music recommendation algorithm based on social network which make full use of tags that with personalized description and information of items in this paper, including:First, we analyses relationships of users, items and tags in music social network Last.fm based, mine latent semantic relationship in user-item-tag, and improve the recommendation accuracy on bipartite node structure similar and random walk with restart. At the same time, we verify the important role of tags in recommendation.Second, we analysis kinds of characteristics of music and introduce the concept of music genome. User-defined tags are defined as free genes in this paper for analyze users’behaviors easier.Last, in the background of mobile services unprecedented development, we proposed personalized mobile music recommendation algorithm based on music genome. Based on the label behavior of users and social tagging, we analyze the characteristics of different music gene and the user’s interest. The interest similarities between different users are used for constructing the neighboring relationship between users. Combined with the two factors, we realized a personalized service on mobile.