基于多模型的铁路货运量预测对比  

Comparison of Railway Freight Volume Forecast Based on Multiple Models

作　　者：燕学博 曹雨 YAN Xuebo;CAO Yu(Fujian University of Technology,Fuzhou 350118,China)

机构地区：[1]福建理工大学,福建福州350118

出　　处：《物流科技》2023年第21期74-78,共5页Logistics Sci-Tech

基　　金：福建省数字城乡融合发展科技经济融合服务平台资助项目。

摘　　要：铁路货运量是我国物流的重要组成部分,也是衡量铁路运力的重要指标。文章以国家统计局公布的2005至2022年铁路货运量指标为标的,基于传统ARIMA模型与LSTM模型提出五种ARIMA-LSTM组合模型。通过实验对比得出结论,任意组合模型拟合效果均优于单一模型,而其中基于方差倒数法的组合模型拟合效果最佳,该模型对比ARIMA模型的MSE指标、RMSE指标、MAPE指标、MAE指标分别降低15.26%、15.62%、24.64%、17.12%,对比LSTM模型分别降低25.32%、32.67%、43.66%、28.33%,经过验证,ARIMA-LSTM组合模型的泛化能力强于单一模型,具有很好的研究与使用价值。Railway freight volume is an important part of China's logistics,and also an important indicator to measure railway transport capacity.Based on the railway freight volume indicators published by the National Bureau of Statistics from 2005 to 2022,this paper proposes five ARIMA LSTM combination models based on the traditional ARIMA model and LSTM model.Through experimental comparison,it is concluded that the fitting effect of any combination model is better than that of a single model,and the combination model based on the reciprocal of variance method has the best fitting effect.Compared with ARIMA model,the MSE index,RMSE index,MAPE index and MAE index of this model are reduced by 15.26%,15.62%,24.64%and 17.12%respectively,and compared with LSTM model,they are reduced by 25.32%,32.67%,43.66%and 28.33%respectively.After verification,ARIMA-LSTM composite model has better generalization ability than single model,and has good research and use value.

关 键 词：货运量 ARIMA LSTM 组合模型 

分 类 号：F530[经济管理—产业经济]