Abstract: The main purpose of burst topic diffusion modeling and prediction is to control the subsequent large-scale dissemination of emergency incidents with adverse effect. Currently microblog topic diffusion and prediction is still in its infancy. The viral infection model, the message propagation model and topic propagation model were deeply studied and a topic diffusion model was proposed based on fans relationship, user activity and influence. By partitioning microblog users into infected users, tangible user and immune user, the relationship between infected and tangible user was analyzed to predict the scale of users which were infected in next time window. Following "internal and external field strength" concept in topic diffusion model, the proportional relationship between them was studied. Based on the scale of the user, topic diffusion prediction algorithms were proposed based on user and node scale respectively. Experiments show that the former can predict diffusion more accurately but with bad time complexity, and the latter node is more suitable for processing large data sets.