Design And Implementation Of Personalized Music Recommendation System

Under the trend of digital transformation,the Internet not only brings big convenience for users to access information resources,but also puts users in a dilemma where they must choose needed data from massive information.Even though traditional search engines have filtered irrelevant information for users,it is still time-consuming to dig out the right information one by one from numerous search results.Both academia and the business agree that recommendation systems are a more effective solutions to this problem.Music is an indispensable form of entertainment in people's daily lives.With the technical improvements of Internet services,typical music consumption behavior has been changed significantly.The user-centric technology has become the most important part of music service provision.Social tags indicate the items' characteristics and also reflect the users' sentiment.Meanwhile users' interest in tags is changing constantly.In this thesis,based on social tags and weighted time,an enhanced music recommending method is proposed.In this method,each user's preference can be modeled according to his listening behaviors,and the songs' features can be represented by their frequency in the attached tags.Then the users' preference degree for each unknown song is measured by the weighted tags' features,and the candidate songs with higher predicted value have the priority to be presented on the recommendation list.Tags contain the arguments for recommended items,which will improve the acceptance of users.To address the sparsity problem,this thesis strengthens the utilization of implicit data.To solve the cold start problem for new users,a solution based on users' demographic information is proposed.Firstly,clustering users with similar behaviors in the system by the similarity calculation rule derived from user-item-tag tripartite graphs diffusion recommendation algorithm;establishing relations between the categories and the users' basic information.Finally,the new users' similar group can be found out according to their basic information.The new user's sentiment can be inferred by similar users,then a personalized recommendation list can be generated.In the end,according to the requirements of the personalized music recommendation system,the overall architecture of the system is designed,each divided functional submodule is described in detail and implemented.Music visualization technologies afford friendly interactive interfaces that intuitively reveal the latent relevance among items,and give users an excellent visual experience...