Leveraging Multiactions to Improve Medical Personalized Ranking for Collaborative Filtering

Algorithm 1

The algorithm of Medical Bayesian Personalized Ranking over multiple users’ actions.

Input: Observed feedback and auxiliary feedback

Output: Parameters

Initialization: for ;; do

   Split n items into three parts: , , ;

end

for iterations do

 for training sample do

   Step 1. Uniformly sample a user ;

   Step 2. Uniformly sample an item i from ;

   Step 3. Uniformly sample an item k from ;

   Step 4. Uniformly sample an item j from ;

   Step 5. Calculate ;

   Step 6. Update via (17), (24);

   Step 7. Update via (18), (25) and the latest ;

   Step 8. Update via (19), (26) and the latest ;

   Step 9. Update via (20), (27) and the latest ;

   Step10. Update via (21), (28);

   Step11. Update via (22), (29);

   Step12. Update via (23), (30);

  end

end