신뢰 파급을 이용한 Top-N 추천 방안

초록

As the vitalization of e-commerce, the importance of recommender systems as a research area and a practical business tool has grown. In order to recommend items to a target user, it is important to rank the items in order of the preference of the target user and pick a specific number of items from them. In this paper, we propose a novel top-n recommendation method probabilistically determines the preference of the target user on each item. We employ Belief Propagation to infer the preference of the target user's preference on each item. We analyze changes of the accuracy of the proposed method according to edge threshold settings and node priors settings while graph construction. Also, by experiments, we show that the proposed method is about maximum 40% more accurate than a existing method based on Random Walk with Restart.

키워드