|
Recommender systems have emerged as a useful tool to help users in discovering personalized items based on their interests.A typical question in recommender system is to predict users' ratings on items based on their history ratings. Through numerousmethods have been proposed to solve this problem, the prediction accuracy is still limited due to the extreme sparsity of the rating data.Therefore, it's necessary to incorporate other information to improve recommendation. To that end, in this paper, we propose a GenerativeNeighborhood-aware Matrix Factorization~(GNaMF) model, which incorporates the tagging data in recommendation. Specially, the tagging datais first used to select neighbors of each user and each item. Then the neighborhood information is incorporated into matrix factorizationin a totally generative way. Compared to other method, this model can more accurately capture each user(item)'s personality,in the meanwhile,the model can learn all the parameters without human effort.Experimental results on the real-world dataset show the improvement of the proposed algorithm |
|
Keywords:Computer Application Technology; Recommender System; Collaborative Filtering |
|