Abstract: 【Purpose/significance】The accurate construction of microblog user portrait can effectively identify the needs of users and
improve the accuracy of personalized recommendation. In order to solve the problem of incomplete extraction of user features, a user portrait construction method is proposed based on interest transfer.【Method/process】Firstly, the weights of different behaviors are de⁃termined by AHP, and used to revise the weights of interest words to obtain the initial interest word set of users; Secondly, the interest behavior cycles of users are obtained according to the life cycle theory, and the time decay function of interest transfer is proposed to dynamic update and superpose the original interest words; Finally, the static attribute labels and the dynamic interest based on interest transfer are integrated to build microblog user portrait.【Result/conclusion】Experiments use the real data crawled from Sina Weibo as data set. The experimental results indicate that compared with the existing Weibo user portrait construction method, the method pro⁃posed in this paper has better performance in personalized recommendation.【Innovation/limitation】The innovation points are: using the life cycle theory to describe the interest behavior cycle of microblog users, constructing the time decay function of interest transfer,and realizing the dynamic update of interest tags. The limitation is that the importance of static attribute tag is not defined, and the in⁃terest behavior curve with abnormal fluctuation is not deeply discussed.