Abstract: Railway freight volume is affected by many factors. Accurate forecasting can provide an important reference for the future planning of the railway industry, and it can also enable the railway sector to formulate transportation policies that are in line with the current freight market. Freight volume data is non-linear and unstable. It is difficult to describe the overall characteristics using traditional single prediction models for prediction, and the prediction accuracy needs to be improved.Based on the principle of decomposition-ensemble, this paper uses the variational modal decomposition algorithm to decompose the freight volume into high-frequency and low-frequency modes. Based on the characteristics of each modal, a prediction model is established, and the obtained prediction results are added up as the forecast value of the final freight volume. The empirical results show that the decomposition-ensemble forecasting method improves the accuracy of forecasting compared with the traditional single forecasting model and can be well applied in the research of railway freight volume demand forecasting.

Key words: railway transportation, railwayfreightvolumeforecasting, decompose-ensemble, variational mode decomposition, ARIMA model, support vector regression

摘要： 铁路货运量受到多种因素影响,准确的预测可以为铁路行业未来规划的编制提供重要的参考依据,也可以使铁路部门制定符合当前货运市场的运输政策。货运量数据具有非线性、不平稳的特点,利用传统的单一预测模型进行预测,很难描述整体特征,预测精度有待提高。本文基于分解—集成的原则,利用变分模态分解算法将货运量分解为高频和低频模态,针对各模态特点,分别建立预测模型,将得到的预测结果加总起来作为最终货运量的预测值。实证表明,分解—集成预测方法与传统的单一预测模型相比,提高了预测的准确率,可以很好地应用在铁路货运量需求预测的研究中。

关键词: 铁路运输, 货运量预测, 分解—集成, 变分模态分解, ARIMA模型, 支持向量回归