基于ARIMA-LSTM-XGBoost组合模型的铁路货运量预测  

Prediction of Railway Freight Volume Based on ARIMA-LSTM-XGBoost Combination Model

作　　者：龙宇 许浩然 余华云[1] 何勇 徐红牛 LONG Yu;XU Hao-ran;YU Hua-yun;HE Yong;XU Hong-niu(School of Computer Science,Yangtze University,Jingzhou 434023,China)

机构地区：[1]长江大学计算机科学学院,荆州434023

出　　处：《科学技术与工程》2023年第25期10879-10886,共8页Science Technology and Engineering

基　　金：国家自然科学基金(61440023);中国高校产学研创新基金-新一代信息技术创新项目(2020ITA03012)。

摘　　要：为提升铁路货运量预测精度和泛化能力,综合考虑铁路货运量时间序列数据的线性和非线性特征,提出了基于ARIMA-LSTM-XGBoost组合模型的铁路货运量预测方法。首先,使用差分整合移动平均自回归(autoregressive integrated moving average, ARIMA)模型对中国铁路货运量进行初步预测;其次,利用长短时间记忆(long short-term memory, LSTM)神经网络对残差进行校正,并将其与极端梯度提升(extreme gradient boosting, XGBoost)模型结合,采用误差倒数法确定权重,构建一种加权组合模型;最后,将组合模型与ARIMA、ARIMA-LSTM、LSTM、XGBoost模型进行对比,借助均方误差(mean square error, MSE)、均方根误差(root mean squared error, RMSE)、平均绝对值误差(mean absolute error, MAE)、平均绝对百分比误差(mean absolute percentage error, MAPE)对上述模型的预测精度进行对比分析。使用2007—2021年全国铁路货运量月度数据进行实验,实验结果表明:组合模型的MSE、RMSE、MAE、MAPE分别为0.011 9、0.109 4、0.068 3、1.775 2%,预测误差均低于上述对比模型,模型的预测精度和泛化能力都有所提升。In order to improve the prediction accuracy and generalization ability of railway freight volume,considering the linear and nonlinear characteristics of railway freight volume time series,a railway freight volume prediction method based on ARIMA-LSTM-XGBoost combined model was proposed.Firstly,autoregressive integrated moving average(ARIMA)model was used to preliminarily predict China s railway freight volume,and then long short-term memory(LSTM)network was used to correct the residuals.Secondly,by combining with extreme gradient boosting(XGBoost)model,a weighted combination model with the weight determined by error reciprocal method was constructed.Finally,the combination model was compared with ARIMA,ARIMA-LSTM,LSTM and XGBoost models,and the prediction accuracy of the above models were analyzed by means of mean square error(MSE),root mean square error(RMSE),mean absolute error(MAE)and mean absolute percentage error(MAPE).Based on the experiment of the monthly data of China s railway freight volume from 2007 to 2021,the results show that the MSE,RMSE,MAE and MAPE of the combined model are 0.0119,0.1094,0.0683 and 1.7752%respectively.The prediction error is lower than the above comparison models,and the prediction accuracy and generalization ability of the model are improved.

关 键 词：铁路货运量预测 ARIMA LSTM XGBoost 组合模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] U294.13[自动化与计算机技术—计算机科学与技术]