Abstract: In order to improve the of efficiency and accuracy of collaborative filtering recommendation, and provide personalized recommendation service to users, a novel collaborative filtering recommendation algorithm based on user score and user attributes similarity is proposed. Firstly, the similarity between the users is calculated according to the similarity of user scores, similarity of the user interest tendency, confidence. Secondly, the similarity between users is measured based on user attributes. Finally, the paper uses MovieLens data set and Book-Crossing data set to do comparative test, such as comparing precision, versatility and performance in different sparsity degree and cold start condition. The result shows that the proposed algorithm not only can improve the recommendation accuracy, but also is better than other collaborative filtering algorithms, and it has higher practical application value.

Key words: recommendation system, collaborative filtering, similarity measurement, sparsity problem