Abstract: Passenger flow forecast is an important basis for railway passenger transport operation and management. The railway passenger flow has the characteristics of non-linear and non-stationary, and the traditional prediction model is difficult to get satisfactory results, so this paper used Empirical Mode Decomposition (EMD) method to decompose the passenger flow adaptively, used Support Vector Regression (SVR) to predict the Inherent Mode function(IMF), and established the SVR passenger flow prediction model based on EMD.The paper used Matlab to simulate SVR prediction, BP neural network prediction and SVR prediction model based on EMD. The average relative errors of the three prediction models were 22%, 25% and 13%, respectively.The simulation results show that the prediction accuracy of SVR method based on EMD is significantly higher than the other two methods, which can effectively improve the accuracy of railway passenger flow prediction.