As the most popular personalized recommendation method, collaborative filtering recommendation has some shortcomings in real application, and the results of recommendation are not accurate enough in some cases. Taking into account the relationship between the trust and similarity of user preferences, we introduce trust into recommendation model, and take dimensional factor indicating user preferences into account, putting forward a recommendation method based on trust and preference. This method is aimed at solving the problem in collaborative filtering method and improving the accuracy of recommendation system. We have demonstrated the effectiveness of this method.