A method and apparatus for predicting content preferences of an active user with reference to content preferences of other users that are similar to those of the active user, the method including: measuring similarity ratios of the active user and other users using an overlapping ratio of use of content items commonly used between the active user and the respective other users; generating a recommender list for the active user based on the measured similarity ratios; and predicting a preference rating of the first user with respect to an unused content item based on a preference rating of an other user included in the recommender list with respect to the unused content item. Accordingly, the similarity ratio between the users can be accurately measured so that reliability of a preference rating prediction system can be increased.