基于EMD-APSO-SVR组合模型的铁路货运量预测

Rail Freight Volume Forecasting Based on Combined EMD-APSO-SVR Model

为了提高铁路货运量预测的准确性,减少类似新冠疫情这样突发事件对现有预测结果的不稳定影响,提出一种基于EMD和APSO优化的SVR组合预测模型。在综合考虑铁路货运量受GDP、铁路营业里程、原煤产量、钢材产量、第二产业增加值以及新冠疫情等因素情况下,对选取的货运量序列进行EMD处理,得到不同时间尺度下的IMF和残差;通过APSO优化SVR模型的参数,并利用优化后的组合模型对各IMF分量分别进行预测,相加得到最终结果。将APSO-SVR模型与EMDAPSO-SVR模型的预测状况进行对比,结果表明,本文建立的EMD-APSO-SVR模型的预测结果误差更小,其预测值与真实值的测定系数高于APSO-SVR模型;前者的平均绝对百分比误差仅有0.22%,能有效提升铁路货运量短期内的预测精度。

In order to improve the accuracy of rail freight forecasting and reduce the unstable impact of unexpected events similar to COVID-19 on the existing forecasting results,a combined prediction model for SVR based on EMD and APSO optimization is proposed.Under the comprehensive consideration of railroad freight volume subject to GDP,railroad business mileage,raw coal production,steel production,secondary industry added value and COVID-19,the selected freight volume series are processed by EMD to obtain IMFs and residuals under different time scales;the parameters of the SVR model are optimized by the APSO algorithm under an adaptive variation strategy;and this combined model is used to predict each IMF component separately and sum up to obtain the final results.The prediction status of the APSO-SVR model is compared with that of the EMD-APSO-SVR model,and the results show that the prediction result error of the EMD-APSO-SVR model established in this paper is smaller and the coefficient of determination of its predicted and true values is higher than that of the APSO-SVR model;the average absolute percentage error of the former is only 0.22%,which can effectively improve the short-term railroad freight volume prediction accuracy.