Research On A Mixed Algorithm Of Song Recommendation Based On Random Walk

The rapid development of Internet technology has resulted in more and more various music websites and applications,fully satisfying people’s needs for music.However,the subsequent information dissemination also brings a huge challenge,that is,it is difficult for users to effectively solve this problem in a large number of music recommendation systems,and to improve the information through personalized recommendation technology to overcome the problem.The sparseness of the data caused by the imbalance of data in the system seriously affects the accuracy of the recommendation results.Therefore,in this paper,the comprehensive similarity of songs is calculated,the approximate correlation map of songs is established,and the probability matrix of field theory theory is integrated to improve the random walk recommendation algorithm.The specific work is as follows:(1)Combining collaborative filtering and content features to build an approximate map of songs.The user-song bipartite graph is established based on the user’s listening record,and the similarity of song scores is calculated based on the number of listening times.The content characteristics of the songs are then extracted to calculate the similarity of song information.Then the two are fused to get the comprehensive similarity,reduce the sparseness of the song similarity matrix,and merge the song priority correlation map.(2)Combine the field theory to establish the transition probability matrix in the random walk algorithm.In calculating the transition probability matrix,the field theory is integrated,and the song field is established from the song correlation map.The importance of the song and the comprehensive similarity are used to obtain the transition probability matrix,thereby reducing the time complexity of random walks.(3)Verification of hybrid recommendation algorithm.Using a subdataset of the millions of songs dataset,a random recommendation song hybrid recommendation algorithm incorporating field theory theory and an item-based collaborative filtering song recommendation algorithm are compared,and a random recommendation song comparison algorithm is compared.The recommendation accuracy rate and recall rate are selected as evaluation indicators to verify the effectiveness of the hybrid algorithm.The experimental results show that the random walking song hybrid recommendation algorithm incorporating field theory theory improves the data sparseness problem to a certain extent and improves the accuracy of the recommendation,which shows that the proposed algorithm is reasonable and effective.