The invention relates to a microblog popularity prediction model based on time and a forwarding sequence of a user, which belongs to the field of message popularity prediction in a social network, andcomprises the following steps of: S1, modeling the forwarding sequence of a microblog by using a recurrent neural network to capture long-distance dependence of a message propagation process; s2, performing nonlinear network transformation on an output result of the hidden layer, and learning the rate of each time step in the propagation process; and S3, utilizing the early trend acceleration andthe early popularity obtained by the rate, and predicting the future popularity of the microblog under the optimization of the user activity. According to the method, it is guaranteed that the popular trend of the microblog in the future is more accurately predicted in the early stage of message propagation, and the model not only utilizes historical propagation information, but also well describes the propagation process of the microblog.