Research On Personalized Music Recommendation System

The popularity of the Internet and the increasing of information resources,the At present,users on the network face a large number of information,it is very difficult to find the information they need quickly and easily.The result of this phenomenon is that the utilization of information is reduced,which is the problem of information overload.In our daily lives,we basically use traditional search engines,such as Baidu,and when we search,we pick out some of the information we need,but there is a lot of redundancy that leads to information overload in people's lives.Information consumers and information creators face a bigger challenge in this era: it's hard to find the information they need on the web,and it's a very real thing for consumers to do.On the contrary,in the face of the huge amount of information appearing on the website,it is difficult for the information producers to effectively present their information to the consumers who need it.The recommendation system is an important tool for resolving this contradiction.The recommendation system meets the needs of information producers and consumers,and can effectively present the information of information producers to the consumers who need it.Nowadays,people are more and more looking for the quality of life,music occupies a large proportion of the entertainment in people's lives,and the rapid development of online music,music recommendation system seems to be more and more necessary.In recent years,music recommendation has also received close attention from domestic and foreign scholars,resulting in a number of research achievements,but also the emergence of many well-known personalized music stations,such as the famous foreign stations Pandora and Million Song The domestic Douban radio station,but its recommended results are low in accuracy and coverage,lack of personalized,many times can not be truly satisfied by users.According to the survey,consumers enjoy listening to music on the air and are less likely to actively seek out their own music preferences.Therefore,this article will be based on the user's behavior log data,combined with the corresponding algorithms to recommend the users to their favorite music.The research of this paper is mainly based on collaborative filtering recommendation algorithm.The data set used in the experiment is Million Song,which provides a public high quality,highly available data set for the research of music information retrieval.In this paper,a brief analysis of the Million Song data set is made,and the characteristics of the user's listening to music are obtained,which provides the basis for the design of the system and the experiment of the algorithm.Then we discuss the way of converting implicit feedback data into scoring,and complete the design of personalized music recommendation system,and achieve some functions.Finally,we do personalized recommendation experiments on data sets.