First, let’s review the recommended features of Cloud Music. Music recommendation is the most direct embodiment of the vision of the founder, Mr. Ding Lei. It is also the main function and core competitiveness of NetEase Cloud Music, and is highly respected by users.
Simply put, the recommendation Buy email list algorithm is to divide the users in the massive user data (behavior records, etc.), and recommend the music that other users like to the same group of users. This requires classifying music and establishing scoring rules, establishing user models, and finding similar users. Classify and match songs based on user behavior data - enabling "blind listening".
NetEase Cloud divides music recommendations into three parts: private FM, daily song recommendations, and recommended playlists.
(1) Analysis from the perspective of accuracy and diversity
Private FM (low accuracy, high diversity): high diversity can bring freshness to users. If you find a song you have never heard but like, it will bring a sense of surprise and mobilize users’ positive emotions. However, due to the low accuracy, it is very likely that the new song is not liked by the user. Therefore, the “Delete” and “Next” buttons are set on the playback interface of the private FM to facilitate the user to switch
Daily song recommendation (high accuracy, low diversity): high accuracy makes the daily recommended 20 songs better meet the user's taste, but there is a problem of single music type, so a playlist is set up to provide users with browsing , the right to operate, make up for the disappointment of the single track to the user.
Recommended playlist (in accuracy, in diversity): The recommended playlist is different from the other two personalized recommendation functions. The threshold of its accuracy and diversity is not only determined by the algorithm, but also determined by its functional form First, the object-oriented functions of the function are divided into two categories, one is the user, and the other is the UGC playlist. The system tags the playlist and the user respectively to improve the accuracy. Since the UGC playlist is created by many users, so UGC playlists have diversity, and the combination of the two ensures the coexistence of accuracy and diversity.