新型冠状病毒肺炎疫情影响下中国航空货运量分析与预测  

An Analysis and Forecasting of Air Cargo Volume in China Under the Impacts of COVID-19 Epidemic

作　　者：陈亚东 丁松滨[1] 刘计民 宋晓敏 隋东[1] CHEN Yadong;DING Songbin;LIU Jiming;SONG Xiaomin;SUI Dong(College of Civil Aviation,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China;East China Regional Air Traffic Management Bureau of Civil Aviation Administration of China,Shanghai 200335,China)

机构地区：[1]南京航空航天大学民航学院,南京211106 [2]中国民用航空华东地区空中交通管理局,上海200335

出　　处：《交通信息与安全》2022年第2期155-162,共8页Journal of Transport Information and Safety

基　　金：中国民用航空局基金项目(TM2019-9-1/2);中国民用航空局华东空管局基金项目(1007KFA21061)资助。

摘　　要：在新型冠状病毒肺炎疫情对航空货运的影响下,月度航空货运量出现异于历史趋势的极端数据,而传统航空货运量预测模型有在极端数据影响下误差较大的问题。因此,研究了适用于后疫情时代的中国航空货运量短期预测方法。对2009-2020年中国航空货运量月度数据进行分析,发现中国航空货运量呈稳定增长趋势。受疫情影响出现短期剧烈波动,在假设疫情对航空货运的影响逐渐减弱的前提下,选取Holt-Winters乘法模型与求和自回归移动平均ARIMA乘积季节模型分别提取航空货运量数据的长期趋势、周期特征和短期波动特征,并采用4种不同权重确定方法构建了多个航空货运量组合预测模型。运用Holt-Winter模型、ARIMA模型及其组合预测模型对2021-2022年中国航空月度货运量进行了预测,以2021年1月—5月的航空货运量数据作为验证数据集,对比分析了不同预测模型的预测误差。结果表明:Holt-Winters与ARIMA组合预测模型的平均绝对百分比误差与最大绝对百分比误差普遍小于自身单一模型的;基于最小二乘法赋权的组合模型预测效果最优,基于残差倒数法赋权的组合模型预测效果次优;最优组合模型的平均绝对百分比误差为1.93%,比次优组合模型降低了8.53%,较单一的Holt-Winters模型与ARIMA模型分别降低了71.70%与20.58%,验证了最优组合模型对后疫情时代中国航空货运量月度数据预测的有效性。With the impacts of COVID-19 epidemic on air cargo market,monthly air cargo volumedata in China shows extreme values,whichare inconsistent with historical trends. As traditional forecasting modelsof air cargo volume are susceptible to large errors due to extreme data,several short-term forecastingmethodsare proposed and developed to forecast air cargo volume in the post-epidemic era of China. It is found thatair cargo volume in China under the influence of COVID-19 epidemic has a steady growth upward trend along with a significant,short-term fluctuation after analyzing the monthly data of air cargo volume in China from 2009 to 2020. Assuming the impactsof COVID-19 epidemic on air cargo decrease gradually,Holt-Winters multiplication model and autoregressive integrating moving average(ARIMA)multiplication seasonal model are applied to model the long-term trend,periodic characteristic,and short-term fluctuation of air cargo quantities respectively. In addition,four different methods for selecting the weights are applied to these two models,in order todevelop combined forecasting models of air cargo volume. Holt-Winter model,ARIMA model,and the combined forecasting model based on the two techniques are used to forecast monthly domestic air cargo volume from 2021 to 2022. The forecasting errors of these models are compared and analyzed based on domestic air cargo volume data from January to May in 2021. The results show that the average absolute percentage error(AAPE)and the maximum absolute percentage error(MAPE)of the Holt-Winters and ARIMA combined model are generally smaller than those of any single model. The combined model weighted by the least square method is found to be most accurate,while itthat based on weights determined by residual reciprocal method is ranked second. The AAPEof the combined model is 1.93%,which is reduced by8.53% whencompared with the combined model ranked second,and is 71.70% and 20.58% lower than that of single Holt-Winters and ARIMA model. Therefore,the effectiveness and accuracy of th

关 键 词：航空运输 航空货运量预测 Holt-Winters乘法模型 ARIMA乘积季节模型 组合预测 新型冠状病毒肺炎疫情 

分 类 号：U8[交通运输工程]