Research On Music Recommendation System Based On Latent Factor Model

With the continuous enrichment of the music resource library,ordinary music search methods can no longer meet the development needs of mass music libraries.The emergence of music recommendation systems can help users solve this problem.Although many music platforms have launched music recommendation services at present,the recommendation algorithm they use cannot fully tap the connection between users and music,and cannot personalize music recommendations based on user preferences.Therefore,this paper proposes a music recommendation system based on latent factor model,which can alleviate the above problems to a certain extent.The recommendation algorithm,model combination and user module constitute the core of the music mixing recommendation system in this paper.The basic idea of the hybrid recommendation system is:first collect the historical behavior of users listening to songs in a specific time period,combine these historical behaviors with a latent factor model,construct a user rating matrix,and then reduce and decompose the rating matrix to finally build the user-Music preference model;at the same time,a time change function is added to the model to consider the user’s preference changes during this time period;finally,by calculating the most similar music to the user,a song recommendation list is generated for the target user.In order to verify the effect of the music recommendation model in the article,a user-music data set suitable for the model in this article was selected.The root mean square error,accuracy rate,recall rate and comprehensive evaluation indicators are used to detect the music recommendation list generated by the model;at the same time,the impact of the accuracy of the prediction score and the accuracy of the recommendation results on the recommendation system in this paper is considered.The final experiment shows that the music recommendation model proposed in the article can improve the accuracy of music recommendation.Compared with the traditional collaborative filtering recommendation algorithm,this paper uses the latent factor model to reduce the dimensionality of the user’s rating data,which can represent the user more accurately The relationship between music alleviates the problem of data sparsity;on this basis,a time function is added to the scoring matrix,which can dynamically display changes in user preferences and recommend more suitable music for users.