Music Recommendation Algorithm Based on Knowledge graph Propagation User Preference

Now a days, researchers have applied auxiliary information to music recommendation algorithms in order to solve the inevitable problems of data sparsity and cold start in recommendation systems, and obtain more potential information through data mining to improve the accuracy of recommendation. This paper describes a model SYT_RippleNet, which combines knowledge graph with deep learning, The knowledge graph is used to explore the potential connection between users and projects and to find the potential interests of users. Then, it promotes the propagation of user preferences on the entity set of the knowledge graph, which is realized through the triple attention mechanism during the preference propagation. Finally, the user's preference distribution for candidate items formed by the user's historical click information is used to predict the final click prediction rate. The music data set Last.FM is applied to SYT_RippleNet model, and good recommendation prediction results are achieved. In addition, the improved loss function is used in the model and optimized by Adam optimizer. Finally, the tanh function is added to predict the click probability to improve the recommendation performance. Compared with the current mainstream recommendation methods, SYT_RippleNet recommendation algorithm has a very good performance in AUC and ACC evaluation indicators, and has a substantial improvement in music recommendation.