Beyond ratings: Alleviating New User Problem in Collaborative Filtering Systems Using Demographic Data

Collaborative filtering is one of the most common techniques in recommender systems. Despite its success and widespread use, it has a problem while facing new users; as recommendations are based on history of user's interactions, system does not know anything about new users' preferences to generate recommendations. Common method to handle this problem is presenting initial list of items to users and get their opinion about items. This paper presents a method based on demographic features of users which in contrast to previous methods, considers users' out of interest items and does not have bias toward a small set of items. We compared this strategy with two other strategies: presenting popular items to new users and random method. Experiments on MovieLens dataset showed that the proposed method gets the most valuable ratings for more items and preserves new user's effort to express his preferences appropriately

کلیدواژه ها:

Recommender systems ، Collaborative filtering ، New user problem ، Demographic features