Collaborative filtering recommendation algorithm based on user interest characteristics and item category

Collaborative filtering is one of the most widely used and successful technology in the personalized recommender system so far. However, because of the influence of data sparsity, the similarity computed based on traditional user-item rating matrix has a lower accuracy, which seriously debases the quality of recommendation systems. In order to alleviate the problem, we take users' interest characteristic information into account besides users' rating information. What's more, we compute one kind of similarity between users based on user-item category rating matrix and another kind of similarity based on user-item category interest separately, and regard the combination of the two kinds of similarity as the final similarity between users. Owing to the above factors, we propose a collaborative filtering recommendation algorithm based on user interest characteristics and item category. The experimental results show that the proposed algorithm can overcome the influence of data sparsity to some extent, and improve the quality of recommendation. © 2013 Binary Information Press.