Design And Implementation Of Music Recommendation Based On Learning To Rank

Music is not only an important part of people's entertainment life,but also an important way to relax.With the rapid development of the Internet,music production technology,music creation cycle and compression technology have improved,and people can easily publish their own music to the Internet,so that more and more musical works present on the Internet.At the same time,Internet is a key medium to obtain resources,which can provide users with a variety of services,so that people get more and more music information through a variety of channels.However,this phenomenon also brings the “ information overload” problem,which makes users pay more time and effort to find their favorite music from the massive music resources.Music recommendation system can find personal preferences based on the interactive records,and then record the music works for users,which can effectively alleviate the "information overload" problem.Content-based recommendation algorithm and collaborative filtering recommendation algorithm are two common recommendation methods in the recommendation system.Music compositions are different from the text information,and it is difficult to record music works based on the music content.To collaborative filtering recommended algorithm,because few users directly score the songs,the users' historical behavior needs to be converted to scores,which always loses a lot of useful information.In order to improve ranking accuracy of music recommendation,this paper studies the music recommendation method based on learning to rank,and designs a music recommendation system based on learning to rank.The system analyzes the user's preference for music according to users' behavior such as downloading,collection,and listening repeatedly.The system provides a sort of product list for users based on the user's product preference and relevance.The main work of this thesis is as follows:(1)This thesis studies several recommendation algorithms and learning to rank algorithms,including comparing the advantages and disadvantages of content-based recommendation algorithm,collaborative filtering recommendation algorithm and hybrid recommendation algorithm,analyzing the application scenarios of three sorting learning methods at point level,pair level and list level,and exploring how to apply the sorting learning method to the recommendation system.(2)In this thesis,Ali music data set provided by "Time Hacker Data Mining" is used as the experimental data set,and the recommended algorithm based on learning to rank is experimented.At the same time,this paper compares the experimental result of the algorithm we proposed with the results of collaborative filtering algorithm based on users,collaborative filtering algorithm based on products,and the non-weight learning to rank method.(3)This thesis designs and implements a music recommendation system based on learning to rank,which is based on the MVC framework and contains four functional modules: recommendation module,login-registration module,record-collection module and song search module,and then test the operation effect of the system.