Abstract: To improve the prediction accuracy and generalization ability of the number of trains operating on the Xian China Railway Express, a prediction method of train number in operation based on the SARIMA-RF combination model was proposed, comprehensively taking into account the linear and nonlinear characteristics of the time series data of the Xian China Railway Express. Firstly, the seasonal autoregressive moving average (SARIMA) model was used to predict the number of vehicles in operation. Secondly, the random forest (RF) model was used to correct the residuals and construct a combination model. Finally, the combination model was compared with ARIMA, SARIMA, RF and XGBoost. The monthly operating data of the Xian China Railway Express from 2014 to 2023 was used to carry out the experiment. The experiment predicts that the number of vehicles operating in 2024 will be 244000 and 267100 in 2025. The comparison results show that the MSE, RMSE, MAE, and MAPE of the combination model are 0.0037%, 0.0610%, 0.0530%, and 3.41%, respectively, which are higher in accuracy than other models.

Key words: traffic and transportation engineering; “Changan” China Railway Express; seasonal fluctuations; SARIMA-RF; residual correction; development strategies

摘要： 为提升西安中欧班列开行车数预测精度和泛化能力，综合考虑西安中欧班列时间序列数据的线性和非线性特征，提出基于SARIMA-RF组合模型的班列开行车数预测方法。首先使用季节性自回归移动平均模型（SARIMA）预测开行车数，其次利用随机森林（RF）模型校正残差，构建组合模型，最后将组合模型与ARIMA、SARIMA、RF、XGBoost进行对比。使用2014—2023年西安中欧班列月度开行数据实验，预测2024年开行车数为24.40万车，2025年为26.71万车，对比结果表明：组合模型的MSE、RMSE、MAE、MAPE分别为0.003 7、0.061 0、0.053 0、3.41%，比其他模型精度更高。

关键词: 交通运输工程；“长安号”中欧班列；季节性波动；SARIMA-RF；残差校正；发展对策