Research And Application Of Music Recommendation System Based On Multi Factor Intervention And Similarity Sensitivity

As we all know,with the rapid development of the Internet,especially the rise of mobile Internet user behavior data generated also rises increasingly,the value of user data generated by both variety and quantity is also more and more,the behavior of the user data,is of great value,recommendation system is in this environment arises at the historic moment,It will combine the user’s information,the information of the item and the user’s preference for the item in the past,use the recommendation algorithm to construct the user interest model reasonably,and provide the user with more effective personalized recommendation service.Traditional music to recommend the most dominant attributes only CARES about music,lyrics,composer and singer,year of issue,play time,etc.,it from two angles of users and items for the inner link between each other,and the research emphasis is more concentrated on the dominant characteristics of the quantitative way and the vector matrix sparse processing,eventually recommend to achieve good effect.The improvement of this paper is mainly carried out from two aspects.First,in terms of recommendation algorithm,theme mining and keyword mining are carried out for music lyrics to form a candidate list of alternative key information.It is hoped to further determine the trend of song style by analyzing the central meaning of lyrics and combining with dominant music attributes.At the same time,in order to avoid errors in semantic analysis of unstructured text,this paper calculates the association relation of elements in the theme and keyword set based on user preferences,and filters out junk key information by setting an empirical threshold to ensure the validity of features.,on the other hand,this paper found that the similarity calculation was carried out on the characteristic matrix,the traditional similarity calculation may be less sensitive data,namely to the absolute value of vector processing ability is weak,and very easy to be Shared between multiple users score item number factors are ignored,which is based on the precision of item recommendation algorithm has made great trouble,So the similarity algorithm is optimized in this paper to solve this problem.Finally,this paper compares the F1 value of the improved scheme and the baseline method,which reflects the positive impact of the improved lyric theme mining strategy proposed in this paper on the recommendation effect.At the same time,in terms of similarity discrimination,the comparison of the effect of the traditional and improved scheme on RMSE evaluation index also verifies the correctness of this idea.