Abstract: Passenger flow prediction is the foundation of the passenger transport organization, and the prediction results can provide the basis for decision making of operation management and emergency disposal. Aiming at the passenger flow prediction of urban rail transit, an autoregressive integral moving average ARIMA model was established for short-term prediction of station inflows, which was based on analyzing the periodic fluctuation law and change trend of the passenger flow in rail transit stations. Taking Beijing subway as an example, an empirical analysis was carried out. The seasonal ARIMA model meeting the requirements was used to carry out the short-term prediction of passenger flow, the actual passenger flow of Dongzhimen station was selected for model parameter calibration, and the passenger flow prediction of typical stations (terminal station, intermediate station, transfer station and interchange station) in the road network was carried out. The research results show that the average prediction error of ARIMA model is only 4%, which has high prediction accuracy and verifies the correctness of the prediction method.

Key words: traffic engineering, urban rail transit, prediction of inflows, ARIMA model, short-term prediction

摘要： 客流量预测是客运组织的基础，预测结果能够为运营管理及应急处置提供决策依据。针对城市轨道交通客流量预测问题，在分析轨道交通站点客流的周期性波动规律及变化趋势的基础上，构建自回归积分滑动平均模型(ARIMA)进行站点进站客流量的短期预测。以北京地铁为例进行实证分析，利用符合要求的季节ARIMA模型对客流量进行短时预测，选取东直门站实际客流进行模型参数标定，并对路网上典型车站(端点站、中间站、换乘站及接驳站)进行客流预测。研究结果表明：自回归积分滑动平均模型的平均预测误差仅为4%，具有较高的预测精度，验证了预测方法的准确性。

关键词: 交通工程, 城市轨道交通, 进站量预测, ARIMA模型, 短期预测