Collaborative filter algorithm is one of the most widely used technologies of personalized recommendation. However, the existing recommendation algorithms only consider the user item rating matrix specific value while calculating the target user neighbor. User preferences and user ratings and the relationship between the project properties are ignored. Moreover, the accuracy also needs to be further improved. To solve this problem, this paper proposed a new collaborative filtering algorithm based on user preferences and project properties (UPPPCF). By using the traditional user project evaluation matrix, the algorithm synthesizes user preferences and the project properties. The project score matrix is changed into project properties score matrix based on user preference. Then the nearest neighbors of target user sets are computed according to this new score matrix. As a result, the proposed algorithm overcomes the insufficiency of existing similarity calculation methods, which only depend on user ratings value. Meanwhile, an effective measurement method for predictor decision is suggested in this paper. The experimental results on MovieLen datasets show that the proposed algorithm can effectively improve the existing traditional collaborative filtering. In addition, the recommendation accuracy has been significantly improved.