The traditional methods of personalized recommendation are faced with the problems of sparse data and cold start. This paper combines the previous research of the project team and introduces the user interest to form the comprehensive similarity, based on the comprehensive consideration of user characteristics and user trust degree. At the same time, this paper uses the social tags which enrich the recommendation data to solve the problem of sparse data in current recommendation system. Firstly, the similarity degree is used to find the similar neighbors of the users and form a tag set by labeling the similar neighbors. Secondly, a tag-based recommendation algorithm is used to generate the final recommendation list. The experimental results show that the proposed algorithm can effectively improve the accuracy of recommendation and the recall rate.