The passenger demand for railway passenger transport directly determines the railway transportation production. It is very important to analyze the historical passenger demand rules accurately and predict the short-term demand in the future. Therefore, the paper proposes a new method of passenger demand analysis based on Prophet model and a short-term passenger flow forecasting method based on Seq2Seq-Attention and Prophet nonlinear combination model. The former can decompose the time distribution characteristics of the passenger flow of the whole data from the long-term historical data, and then analyze the passenger demand law in the past; the latter uses neural network to make a nonlinear combination to give full play to the respective advantages of Seq2Seq-Attention network and Prophet model on different scale data sets, and make more accurate passenger flow demand prediction. The results of the example show that the error of the whole data set is only 6.68% after the passenger flow history data is decomposed into multiple time distribution data by prophet model. At the same time, the prediction effect of Seq2Seq-Prophet model on the data set is better than that of the single model and other existing methods.