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

陈亚东; 丁松滨; 刘计民; 宋晓敏; 隋东

doi:10.3963/j.jssn.1674-4861.2022.02.018

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

doi: 10.3963/j.jssn.1674-4861.2022.02.018

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

基金项目:

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

详细信息

作者简介:
陈亚东（1997—），硕士研究生. 研究方向：航空运输. E-mail: chenyadong1997@163.com

通讯作者:
丁松滨（1964—），博士，教授. 研究方向：民航安全、飞行性能工程研究. E-mail : dingzhili@nuaa.edu.cn

中图分类号: U8
计量
- 文章访问数: 946
- HTML全文浏览量: 438
- PDF下载量: 151
- 被引次数: 0
出版历程
- 收稿日期: 2021-08-21
- 网络出版日期: 2022-05-18

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

1.
College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China
2.
East China Regional Air Traffic Management Bureau of Civil Aviation Administration of China, Shanghai 200335, China

摘要

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。
- 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
Abstract: 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 by 8.53% 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 the optimized, combined model in forecasting the monthly domestic air cargo volume within the post-epidemic era has been verified.
- air transport /
- forecast of air cargo volume /
- Holt-Winters multiplication model /
- ARIMA multiplicative seasonal model /
- combination forecast /
- COVID-19 epidemic

HTML全文

图 1 2009—2020年中国航空货运量月度数据时序图

Figure 1. Time sequence diagram of monthly data of Chinese air cargo volume from 2009 to 2020

下载: 全尺寸图片幻灯片

图 2 Holt-Winters乘法模型预测结果

Figure 2. Prediction results of Holt-Winters multiplication model

下载: 全尺寸图片幻灯片

图 3 运输量一阶差分、一阶季节差分图

Figure 3. 1^st difference and 1^st seasonal difference diagram of transportation volume

下载: 全尺寸图片幻灯片

图 4 运输量二阶差分、一阶季节差分图

Figure 4. 2^nd difference and 1^st seasonal difference diagram of transportation volume

下载: 全尺寸图片幻灯片

图 5 差分处理后自相关图

Figure 5. Autocorrelation diagram after data processing

下载: 全尺寸图片幻灯片

图 6 差分处理后偏自相关图

Figure 6. Partial autocorrelation diagram after data processing

下载: 全尺寸图片幻灯片

图 7 ARIMA(1, 2, 2)(0, 1, 2)₁₂ 模型预测结果

Figure 7. Forecasting results of ARIMA(1, 2, 2)(0, 1, 2)₁₂ model

下载: 全尺寸图片幻灯片

表 1 2009—2020年中国航空货运量月度数据

Table 1. Monthly data of Chinese air cargo volume from 2009 to 2020 单位: 10 000 t

年份	月度货运量
年份	1	2	3	4	5	6	7	8	9	10	11	12
2009	26.43	25.64	33.83	35.21	34.13	33.72	35.99	39.08	45.65	41.83	45.00	45.54
2010	45.07	34.27	47.96	46.60	46.73	43.63	44.54	46.32	51.57	49.64	49.82	51.45
2011	49.65	30.07	48.90	47.83	44.74	43.51	45.29	45.53	50.65	46.94	48.59	49.92
2012	35.42	37.77	46.84	44.53	44.99	42.86	43.73	45.56	52.79	45.62	51.40	50.01
2013	47.66	29.31	48.72	46.17	46.78	43.76	43.97	46.31	52.23	48.80	53.00	51.00
2014	46.80	33.40	51.73	49.47	50.39	46.17	47.19	49.56	53.93	52.96	55.53	55.00
2015	51.86	40.69	52.00	52.22	52.89	49.39	48.72	50.36	55.48	54.17	58.71	58.80
2016	58.80	35.20	56.40	54.10	55.20	53.60	52.40	53.30	59.70	59.00	64.20	65.10
2017	56.50	41.80	58.90	57.50	59.90	58.10	54.80	57.10	64.40	59.00	66.70	66.10
2018	63.90	45.60	61.50	60.50	62.60	61.40	59.10	60.70	66.30	63.10	66.10	67.00
2019	67.20	37.60	63.10	60.80	62.40	60.70	61.60	63.00	69.00	66.50	70.00	71.30
2020	60.60	29.70	48.40	48.40	54.90	57.80	55.20	54.90	66.50	62.10	67.50	69.50

下载: 导出CSV

表 2 Holt-Winters乘法模型统计

Table 2. Statistical test of Holt-Winters multiplication model

模型拟合度统计			杨-博克斯Q(18)检验
平稳R²	R²	MAPE /%	统计	DF	显著性
0.625	0.908	4.536	20.285	15	0.161

下载: 导出CSV

表 3 2021年和2022年航空货运量预测值（Holt-Winters乘法模型）

Table 3. Forecasting results of air cargo volume in 2021 and 2022 (with Holt-Winters multiplication model)

年份	月份	预测值/x10⁴t	年份	月份	预测值/x10⁴t
2021	1	61.92	2022	1	64.65
	2	39.75		2	41.50
	3	62.37		3	65.12
	4	61.79		4	64.51
	5	64.52		5	67.35
	6	62.33		6	65.06
	7	60.38		7	63.02
	8	61.78		8	64.47
	9	70.11		9	73.16
	10	66.34		10	69.22
	11	71.17		11	74.25
	12	71.54		12	74.64

下载: 导出CSV

表 4 备选模型的检验

Table 4. Verification of alternative models

模型参数	R²	正态化BIC
(1, 2, 2)(0, 1, 2)₁₂	0.928	2.276
(1, 2, 2)(1, 1, 2)₁₂	0.928	2.317
(1, 2, 1)(0, 1, 2)₁₂	0.906	2.450
(1, 2, 1)(1, 1, 2)₁₂	0.919	2.440
(2, 2, 1)(0,1,2)₁₂	0.927	2.333
(2, 2, 1)(1, 1, 2)₁₂	0.922	2.401
(2, 2, 2)(0, 1, 2)₁₂	0.950	2.281
(2, 2, 2(0, 1, 2)₁₂	0.908	2.556

下载: 导出CSV

表 5 ARIMA(1, 2, 2)(0, 1, 2)₁₂ 模型统计

Table 5. Statistical test of ARIMA (1, 2, 2)(0, 1, 2)₁₂ model

模型拟合度统计			杨-博克斯Q(18)检验
平稳R²	R²	MAPE /%	统计	DF	显著性
0.921	0.928	3.248	10.215	13	0.676

下载: 导出CSV

表 6 2021年和2022年的航空货运量预测值（ARIMA(1, 2, 2)(0, 1, 2)₁₂）

Table 6. Forecasting results of air cargo volume in 2021 and 2022 (ARIMA (1, 2, 2)(0, 1, 2)₁₂)

年份	月份	预测值/xl0⁴t	年份	月份	预测值/xl0⁴t
2021	1	65.64	2022	l	70.7
	2	49.37		2	55.23
	3	65.79		3	71.37
	4	64.24		4	69.6
	5	66.62		5	72.35
	6	65.59		6	71.78
	7	64.39		7	70.28
	8	65.71		8	71.28
	9	72.82		9	79.16
	10	69.73		10	75.94
	11	74.34		11	80.76
	12	75.38		12	82.14

下载: 导出CSV

表 7 组合模型权重分配

Table 7. Weight distribution of combined model

模型	赋权方法	Holt-Winters乘法模型	arima乘积季节模型
组合模型l	等权平均法	0.50	0.50
组合模型2	方差倒数法	0.42	0.58
组合模型3	残差倒数法	0.39	0.61
组合模型4	最小二乘法	0.30	0.70

下载: 导出CSV

表 8 各模型预测结果对比

Table 8. Comparison of forecasting results of all models 单位: 10 000 t

时间	实际值	arima乘积季节模型	Holt-Winters乘法模型	组合模型1	组合模型2	组合模型3	组合模型4
1月	66.90	65.64	61.92	63.78	64.08	64.19	64.53
2月	45.90	49.37	39.75	44.56	45.33	45.62	46.51
3月	65.50	65.79	62.37	64.08	64.35	64.46	64.77
4月	65.50	64.24	61.79	63.02	63.21	63.28	63.51
5月	66.40	66.62	64.52	65.57	65.74	65.80	66.00

下载: 导出CSV

表 9 各模型误差比较

Table 9. Comparison of model errors

模型名称	MAPE /%	最大绝对百分比误差/%
ARIMA乘积季节模型	2.43	7.56
Holt-Winters乘法模型	6.82	13.40
组合模型1	2.96	4.66
组合模型2	2.34	4.22
组合模型3	2.11	4.05
组合模型4	1.93	3.54

下载: 导出CSV

表 10 中国航空货运量预测结果（组合模型4）

Table 10. Forecasting results of air cargo volume in China (combined model 4)

年份	月份	预测值/x10⁴t	年份	月份	预测值/x10⁴t
2021	1		2022	1	68.89
	2			2	51.11
	3			3	69.50
	4			4	68.07
	5			5	70.85
	6	64.61		6	69.76
	7	63.19		7	68.10
	8	64.53		8	69.24
	9	72.01		9	77.36
	10	68.71		10	73.92
	11	73.39		11	78.81
	12	74.23		12	79.89

下载: 导出CSV

参考文献(20)

[1]	赵瑜. 航空货运为何"一舱难求"?[N]. 中国民航报, 2021-08-12(5). ZHAO Y. Why is air cargo "difficult to find cargo space"?[N]. China Civil Aviation News, 2021-08-12(5). (in Chinese)
[2]	GUDMUNDSSON S V, CATTANEO M, REDONDI R. Forecasting temporal world recovery in air transport markets in the presence of large economic shocks: The case of COVID-19[J]. Journal of Air Transport Management, 2021(91): 102007.
[3]	周叶, 肖灵机. 基于ARIMA模型的中国航空货运量预测分析[J]. 南昌航空大学学报(社会科学版), 2010, 12(3): 22-27. doi: 10.3969/j.issn.1009-1912.2010.03.004 ZHOU Y, XIAO L J. Prediction and analysis of China's air cargo volume based on ARIMA model[J]. Journal of Nanchang Aviation University(Social Science Edition), 2010, 12(3): 22-27. (in Chinese) doi: 10.3969/j.issn.1009-1912.2010.03.004
[4]	文军. 基于灰色马尔可夫链模型的航空货运量预测研究[J]. 武汉理工大学学报(交通科学与工程版), 2010, 34 (4): 695-698. doi: 10.3963/j.issn.1006-2823.2010.04.013 WEN J. Research on air freight volume forecast based on grey Markov chain model[J]. Journal of Wuhan University of Technology(Transportion Science & Engineering), 2010, 34(4): 695-698. (in Chinese) doi: 10.3963/j.issn.1006-2823.2010.04.013
[5]	朱倩, 廖志高, 张峰祎. 基于聚类算法和ANFIS的广西航空货运量预测研究[J]. 武汉理工大学学报, 2015, 37(8): 37-41. https://www.cnki.com.cn/Article/CJFDTOTAL-WHGY201508009.htm ZHU Q, LIAO Z G, ZHANG F Y. Research on Guangxi air freight volume forecast based on clustering algorithm and ANFIS[J]. Journal of Wuhan University of Technology, 2015, 37(8): 37-41. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-WHGY201508009.htm
[6]	汤银英, 李龙. 基于Holt-Winters模型的铁路月度货运量预测研究[J]. 交通运输工程与信息学报, 2017, 15(2): 1-5+13. doi: 10.3969/j.issn.1672-4747.2017.02.001 TANG Y Y, LI L. Research on railway monthly freight volume prediction based on Holt winters model[J]. Journal of Transportation Engineering and Information, 2017, 15(2): 1-5+13. (in Chinese) doi: 10.3969/j.issn.1672-4747.2017.02.001
[7]	LIU J M, DING L, GUAN X Y, et al. Comparative analysis of forecasting for air cargo volume: Statistical techniques vs. machine learning[J]. Journal of Data, Information and Management, 2020, 2(4): 243-255. doi: 10.1007/s42488-020-00031-1
[8]	MANCUSO A C B, Werner L. A comparative study on combinations of forecasts and their individual forecasts by means of simulated series[J]. Acta Scientiarum Technology, 2019, 41(1): 41452. doi: 10.4025/actascitechnol.v41i1.41452
[9]	BLANC S M, SETZER T. Bias-Variance trad-off and shrinkage of weights in forecast combination[J]. Management Science, 2020, 66(12): 1-18.
[10]	SALISU A, SHABRI A, ISMAIL Z. A combine Holt-Win-ters and support vector machines models in forecasting airlines seasonal time series data[J]. Mathematics and Statistics Journal, 2015, 1(4): 8-16.
[11]	TAMBER A J, MICHAEL O O, Ojowu O J. The Holt-Win-ters multiplicative model of passengers' traffic forecast of the Nigeria airports[J]. International Journal of Electrical and Computer Systems, 2021, 3(1): 35-40.
[12]	YANG Y M, YU H, SUN Z. Aircraft failure rate forecasting method based on Holt-Winters seasonal model[C]. 2017 IEEE 2^nd International Conference on Cloud Computing and Big Data Analysis(ICCCBDA), Chengdu, China: IEEE, 2017.
[13]	SIAMI-NAMINI S, TAVAKOLI N, NAMIN A S. A comparison of ARIMA and LSTM in forecasting time series[C]. 2018 17^th IEEE International Conference on Machine Learning and Applications(ICMLA), Orlando, USA: IEEE, 2018.
[14]	陈希睿. 基于时间序列分析的CPI预测[D]. 北京: 清华大学, 2019. CHEN X R. CPI prediction based on time series analysis[D]. Beijing: Tsinghua University, 2019. (in Chinese)
[15]	DUBEY A K, KUMAR A, GARCÍA-DÍAZ V, et al. Study and analysis of SARIMA and LSTM in forecasting time series data[J]. Sustainable Energy Technologies and Assess-ments, 2021(47): 101474
[16]	薛艳茹. 基于时间序列分析的散杂货港口吞吐量短期预测研究[D]. 北京: 北京交通大学, 2019. XUE Y R. Research on short-term forecast of bulk cargo port throughput based on time series analysis[D]. Beijing: Beijing Jiaotong University, 2019. (in Chinese)
[17]	GUO Y H, SHI X P, ZHANG X D. A study of short term forecasting of the railway freight volume in China using ARIMA and Holt-Winters models[C]2010 8th International Conference on Supply Chain Management and Information, Hong Kong, China: The Hong Kong Polytechnic University, 2010.
[18]	詹英. 组合预测方法在中国人均GDP预测中的应用[D]. 武汉: 华中师范大学, 2014. ZHAN Y. Application of combination forecasting method in China's per capita GDP forecast[D]. Wuhan: Huazhong Normal University, 2014. (in Chinese)
[19]	中国民用航空局. 月度生产运输统计[R/OL]. (2021-12)[2022-01-04]. http://www.caac.gov.cn/XXGK/XXGK/index_172.html?fl=11. Civil Aviation Administration of China. Monthly production and transportation statistics[R/OL]. (2021-12)[2022-01-04]. http://www.caac.gov.cn/XXGK/XXGK/index_172.html?fl=11. (in Chinese)
[20]	易丹辉. 统计预测: 方法与应用[M]. 北京: 中国人民大学出版社, 2014: 212-267. YI D H. Statistical prediction: methods and applications[M]. Beijing: China Renmin University Press, 2014: 212-267. (in Chinese)