Research On Music Recommendation Algorithm Based On Fusion Model

With the rapid development of the online music industry,each platform generates massive amounts of music data every day.Faced with a rich library of music,it is difficult for people to quickly find the songs they like.Therefore,there is an urgent need for an effective music recommendation system to help them retrieve music.For users,the music recommendation system can help enhance their experience.For the platform,it enhances the user's stickiness and brings more power to the development of the platform.Some traditional algorithms in the recommended system domain can solve this information overload problem to a certain extent,but the limitations of the algorithm and the particularity of the music still bring great challenges to the recommendation.In order to solve these problems and improve the recommendation effect of music,this paper proposes a fusion algorithm framework.The model fusion method overcomes the limitations of the single model.After enhancing the multi-information of the model,the content-based modeling method is used to predict the user's preference.After the feature engineering and hyper parameter optimization on the dataset,the effect of the algorithm framework is verified by experiments.A prototype system was designed and implemented based on the algorithm framework.The main work of this paper includes the following points:(1)We Analyze the research background,current situation and significance of music recommendation,and put forward the research content and solution of this paper in view of some problems in the current research status.(2)We Analyze related research in the field of recommendation systems,including collaborative filtering,context awareness and content-based recommendation algorithms,which laid the foundation for the research work of this paper.(3)In view of the current data sparseness and low feature utilization,based on lifting tree and logistic regression,we propose two fusion algorithm frameworks and designe feature engineering and hyper parameter optimization.Experiments are carried out on real data.It shows that the proposed fusion algorithm framework solves the limitation of the single model to a certain extent and improves the accuracy of the user's music preference prediction.(4)Based on the proposed fusion algorithm framework,we complete the development of the music recommendation prototype.