Personalized Music Recommendation System Based On Improved Collaborative Filtering

With the rapid development of the Internet and digital music,various music platforms provide a large number of songs for users to listen to.However,as the number of songs increases,it is difficult for users to find interesting music quickly in the face of massive song information.In order to enhance the competitiveness of the industry,more and more music platforms adopt recommendation systems to provide users with high-quality personalized recommendation services.It can help users find their favorite songs quickly,provide users with a good experience,and add user’s satisfaction and loyalty to the music platform.Therefore,the personalized music recommendation system has become the research direction of scholars and the industry.Collaborative filtering algorithms are widely used in recommendation systems because of their simple implementation and strong versatility.However,collaborative filtering recommendation algorithms face many problems in the application process.The most typical problems include cold start,data sparsity and grey-sheep problem.In this paper,some improvements and innovations are explored in the music recommendation system for the traditional collaborative filtering algorithm.Firstly,aiming at the problems of user and item cold start and data sparsity in the music recommendation system,a personalized music recommendation algorithm based on improved collaborative filtering is proposed.Based on the traditional collaborative filtering algorithm,the algorithm combines user portrait,user ratings and music tag data,it uses user portrait to calculate user similarity,which solves user cold start problem,and uses music tag score to initialize unknown ratings to solve music cold start and data sparsity issues.Based on the enrichment of the ratings data,the item-based collaborative filtering method is used to mine user preferences.Secondly,aiming at the problem of gray-sheep user and ratings sparsity in the music recommendation system,a music recommendation algorithm based on collaborative filtering and playing coefficients is proposed.Calculate the user’s ratings on the singer by user’s playing information and frequency linear function,and solve the user rating sparsity problem: first it calculates the singer’s listening coefficient,then calculate the user’s playing coefficient by the singer’s listening coefficient,and use the coefficient as the user attribute,then use the cosine similarity formula to calculate the user similarity.In the end,it adopts the user-based collaborative filtering algorithm to mine the user’s preferences,thus the algorithm solves the problems of gray-sheep users and data sparsity.On the Last.fm music dataset,the proposed personalized music recommendation algorithm based on improved collaborative filtering and the music recommendation algorithm based on collaborative filtering and playing coefficients are evaluated by experiments to verify its effectiveness.The experimental results show that compared with traditional collaborative filtering algorithm,the Root Mean Square Error of the music recommendation algorithm based on improved collaborative filtering is smaller.The music recommendation algorithm based on collaborative filtering and playback coefficients is evaluated in various evaluation standards(RMSE,MAE,NMAE,MAP,NDCG,AUC)are superior to other collaborative filtering methods and their variant methods,which can effectively improve the quality of music recommendation.