Collaborative filtering recommendation algorithms are the most popular approaches in the area of recommender systems and have been extensively discussed by researchers. In this paper, we focus on the analysis of items influence received from neighborhood and the corresponding iterative preference prediction based on the influence. Specifically speaking, the proposed approach uses influence coefficient to measure an item’s ability to influence neighbors’ acceptance by users, and predicts a user’s preference for an item based on the user’s ratings on these items which have influence on the target item. In the meanwhile, the proposed approach distinguishes influence into persuasive influence and supportive influence, and takes into account the combined effect of the two types of influence. Under this methodology, we verified that the proposed algorithm obviously outperforms standard collaborative filtering methods through 5-fold cross validation.
各種検索
サポート
ヘルプ