Design And Implementation Of Personalized Music Recommendation System Based On Spark

In the development trend of Internet,music undoubtedly plays a very important role.The major music platforms provide consumers with countless musical works,so that people can immerse themselves in the ocean of music at all times.However,as Internet resources become more abundant,massive amounts of data often make it difficult for users to find the music they really need.An excellent music recommendation system can solve this problem very well.While recommending suitable songs to users,it can also attract users to the platform and increase commercial benefits.Recommendation technology is becoming more and more mature.The mainstream recommendation technology includes collaborative filtering recommendation technology and content-based recommendation technology.These technologies can recommend items suitable for users or users like in a large amount of data.However,with the development of the Internet and the explosive growth of data,the traditional recommendation system cannot meet the needs of computing under big data,and the accuracy is not optimistic.This thesis has conducted an in-depth study of the music recommendation system,and developed a recommendation system using the Spark platform as the computing engine to solve the music recommendation problem in the big data environment.This thesis designs and implements two recommendation algorithms on the Spark platform: cluster-based ALS model recommendation algorithm and label-based graph recommendation algorithm.The former solves the cold start problem of new users of the recommendation system through parallel mixing of Spark MLlib's ALS recommendation algorithm and Kmeans clustering algorithm,and improves the calculation efficiency and reduces the calculation time by deeply optimizing the Spark calculation process;the latter is based on the Personal Rank algorithm Based on the Spark platform,the music tag is used to implement a music recommendation algorithm based on a graph model,and the TF-IDF(word frequency-inverse document frequency)algorithm is used to solve the long-tail distribution problem existing in music and tags to optimize the recommendation effect.Finally,this thesis builds a Spark computing cluster,designs the underlying storage,designs the front-end and back-end interactions based on Spring Boot and Vue framework,and connects the recommendation engine with the underlying system and Java Web to create a complete interactive personalized music recommendation system.After testing and evaluation,the system can meet user needs and has a good recommendation effect.