Predicting Preference Tags to Improve Item Recommendation.

摘要: Collaborative filtering (CF) based recommender systems identify and recommend interesting items to a given user on the user’s past rating activity. These improve their recommendations by identifying preferences item related information from external sources, like reviews written users, or concept tags shared users about these items. are often reflected through multi-criterion rating. In this study, we seek integrating as side within standard neighborhoodbased matrix factorization methods. We assume that choice of for an provides additional personal preference features item. Since, querying provide multi-criteria imposes burden base, propose using collective classification predict both also investigate use active learning approaches integrated framework when tag (users items) is limited. Our experimental results several real world datasets show advantages tag-based systems. able effectiveness algorithms in estimating features. Keywords— Tag-based Recommendation System, Active Learning, Collective Classification