Research And Implementation Of Music Recommendation Algorithm Based On Social Relations And Item Characteristics

It is difficult for users to retrieve music resources due to information overload,in order to solve the problem,the recommender system has been widely used in the field of music information retrieval.However,the traditional recommendation strategy does not take into account the richness of music features and the diversity of user behavior.Based on the implicit feedback information such as the trust relationship,the music tag and the number of playing times in the music community,the relevant recommendation algorithm is studied in this paper.The main research contents of this paper include the following four parts:(1)A music recommendation algorithm combining trust relationship and user preference is proposed.Trust is one of the most important attributes in the modern social network,and it turns out that users tend to be more inclined to choose what the users they trust like.But there’s a problem,the users they trust may have different preference,and the users they may not trust could have the same interest.Therefore,this paper makes a weighted fusion of the trust relationship and the user’s preference similarity to better excavate the user preference characteristics.(2)An improved music recommendation algorithm based on user topic model is proposed.Music tags contain rich potential semantics,and different tags may represent the same attributes.It is rough to calculate the similarity between the users or the items by simply using the tag co-occurrence.In this paper,in order to fully excavate the distribution of the users and songs in the topic space,and find more similar neighbors for the target users,the topic model is used to do the cluster analysis of music tags,and the number of music plays is used to revise the weight of the topic.(3)A music recommendation algorithm based on the improvement of classified user behavior features is proposed.User preferences is divided into 3 types of user-driven behavior,which not only considers the user’s playback and tag preferences,but also fills the missing user data according to the popularity.A UIP-Walk recommendation algorithm combining user interest,music tag,and popularity is proposed,which uses the improved random walk strategy to calculate the user’s score of the songs,and to evaluate the potential relationship between the user and the music.(4)A music recommender system based on social network trust and text tag is designed and implemented,which includes user center,music playback and music recommendation.Based on the three research findings mentioned above and also considering the social network trust and user topic similarity,using classified user behavior features to alleviate cold start,and improve the recommendation performance.