Abstract: Nowadays, there are many models for the prediction of Weibo propagation, but the factors are not completely comprehensive. To solve this problem, this paper proposes a multifeature prediction model of Weibo propagation based on triplet neural network. The basic framework of this model was a triplet neural network structure. In this model, LDA model was used to extract the text features of microblog, and improved PageRank algorithm was used to analyze the characteristics of user influence. The model combined with other features such as whether the microblog had pictures, links and videos. Experimental results show that the proposed model significantly improves the accuracy of Weibo propagation prediction compared with twobranch models, which has good stability.