Design And Implementation Of Music Recommendation System Based On Big Data Platform

With the development of network technology,music recommendation system has developed rapidly,and online music platform has become the first choice for people to listen to music.However,the music recommendation system also faces some problems,such as chaotic data storage,low computational efficiency,cold start and sparse data caused by large data scale.In view of the above problems,this thesis first summarizes the development status and relevant theories of music recommendation system at home and abroad;Then the recommendation system is analyzed and studied,and a hybrid recommendation algorithm based on collaborative filtering is designed;Secondly,the overall architecture of the music recommendation system is designed,and each functional module is designed and implemented;Finally,the music recommendation system is tested to verify the feasibility and stability of the recommendation system.The main research contents of this thesis are as follows:(1)Improve the item-based collaborative filtering recommendation algorithm.In order to alleviate the problems of cold start and data sparse,the hybrid similarity formula was designed by combining Pearson similarity and label similarity,and then the similarity between music was calculated.In order to capture the weights of the two similarity degrees dynamically,Jaccard dynamic coefficient is added into the mixed similarity formula.Considering the timeliness of user behavior,the time weighting factor is added into the user prediction scoring formula.Finally,the accuracy rate and recall rate are used as evaluation indexes.When the time weight factor is 0.7,the two indexes have the best effect.In addition,the mixed similarity formula based on collaborative filtering is about 2% higher than the other two similarity formulas,which verifies the effectiveness of the improved recommendation algorithm.(2)Design and implement the music recommendation system.To solve the problem of insufficient storage and computing capacity of recommendation system caused by increasing data.Firstly,two channels of offline data transmission and realtime data transmission are designed to collect and transmit data.The music data warehouse is built,and the data is processed and stored hierarchically.Then the data are preprocessed,which is beneficial to the calculation of recommendation model.Secondly,based on the improved algorithm,combined with the distributed framework of Hadoop,the recommendation module is completed and the music recommendation system is implemented.Finally,through functional test and non-functional test,it reflects the efficiency,expansibility and stability of the music recommendation system,which can meet users' personalized music needs.