Abstract: 【Purpose/significance】Aiming at the shortcomings of existing friend recommendation methods in user dynamic interest mining and trust analysis, a friend recommendation method combining user dynamic interest and social trust is proposed【. Method/pro⁃ cess】Firstly, the latent dirichlet allocationmodel is used to learn the topic distribution of the user 's published blog posts, and the time decay function is designed to be weighted to mine the user 's dynamic interest preference characteristics and calculate the similarity of interest preference between users. Secondly, the social network structure features are used to calculate the global trust degree, and the user 's social behavior features are used to calculate the local trust degree, and the final social trust degree is obtained by linear weighting. Finally, the user 's interest preference similarity and social trust degree are fused to recommend potential friends of the tar⁃ get user and generate a potential friend recommendation list【. Result/conclusion】The experimental results based on Sina Weibo data⁃ set show that the proposed method is superior to the existing representative recommendation methods in terms of recommendation ac⁃ curacy and Top-K recommendation ability【. Innovation/limitation】Compared to existing method, this paper comprehensively considers the information of user dynamic interest and social trust to recommend potential friends of the target user. But there is insufficient pa⁃ rameter tuning, and many to many, one to many group trust relationships also need to be fully utilized.