Abstract: To tackle the data sparse problem of the traditional collaborative filtering algorithm, this paper proposes a collaborative filtering algorithm based on item attribute preference (CFBIAP). This algorithm calculates user similarity based on item attribute preference by using the item attributes and scores. Meanwhile it makes linear fitting with the similarity based on score matrix to get the user similarity. Up to a point, it decreases the error merely by score matrix partly. Experiments on MovieLens dataset show that the recommendation is better than traditional collaborative filtering algorithm both in quality and effect. The algorithm solves the data sparse problem effectively.


Key words: collaborative filtering, recommendation, item attribute, similarity