关键词: 铁路运输, 短时客流预测, ＳＡＲＩＭＡＬＳＴＭ组合模型, 滚动优化算法, 注意机制

Abstract: Aiming at the huge disturbance caused by sudden public health events such as the COVID-19 to the short-term railway passenger flow, this paper constructs a combination model based on SARIMA-LSTM to analyze the daily passenger flow curve of the periodic and seasonal non-stationary time series during Spring Festival transportation under the epidemic situation.The SARIMA model is used to predict the linear part, and the LSTM rolling optimization model is used for nonlinear prediction. Finally, the two prediction results are put into the weighted sum of the attention mechanism module, and the GRU gated loop unit is introduced to assist the verification. The analysis shows that the prediction results of SARIMA-LSTM combination model have good control and high accuracy, which can provide theoretical basis for the prediction of the short-term passenger flow data set of epidemic emergencies.