基于时空注意力机制的应急医疗物资需求预测建模研究

doi:10.16381/j.cnki.issn1003-207x.2024.0769

Abstract

Abstract:

The outbreak of public health emergencies， such as COVID-19， necessitates accurate predictions of epidemic trends and medical resource demands to enable timely interventions. Existing methods for epidemic forecasting often fail to simultaneously capture spatial dependencies and temporal dependencies. Additionally， dynamic demand prediction for emergency medical supplies， classified as consumable supplies and non-consumable supplies， remains to be unexplored， particularly under rapidly evolving scenarios. Therefore， it tries to address these gaps for this study by proposing a novel framework for spatiotemporal epidemic prediction and resource demand modeling. A SpacetimeNet is introduced， which is an autoregressive model integrating spatiotemporal attention mechanisms to jointly learn spatial correlations and temporal dependencies. The framework comprises：（1） Spatial Encoder： It employs multi-head self-attention to model dynamic inter-regional influences. For target region $x t i (i ∈ N, t ∈ N)$ ， the spatial attention vector $A t t e n t i o n i (Q i, K, V)$ is computed by $A t t e n t i o n i (Q i, K, V) = s o f t m a x (Q i K T / d k) V$ . （2）Temporal Encoder： It combines positional encoding with temporal attention to capture sequential patterns. （3）Autoregressive Decoder： It uses masked self-attention to iteratively predict future values while preventing information leakage. The final prediction $y^t i$ is derived via $y^t i = R e L u (W p O i' + b p)$ .The used dataset includes 13 features （e.g.， confirmed cases， deaths， vaccination rates） from 5 representative countries （France， Germany， India， USA and UK）. Data preprocessing involves log-transformation and normalization to enhance stationarity. An 80：20 train-test split ensures robust evaluation. Daily new cases are forecasted using the SpacetimeNet， and then actual active cases are computed and medical resource demands are derived based on mortality/recovery rates and staffing ratios.Evaluated on the dataset described， the SpacetimeNet outperformed benchmarks （ARIMA， LSTM， GRU， etc.） with lower MAE， lower RMSE and lower SMAPE. The final results show that in the early stage of the rapid development of the epidemic， the demand for consumable emergency medical supplies rises rapidly， while the demand for non-consumable emergency medical supplies fluctuates considerably， after which the rise in the demand for consumable emergency medical supplies gradually levels off， and the demand for non-consumable emergency medical supplies gradually decreases due to the accumulation of resources to 0. The results of the present study will provide a precise prevention and control of epidemics and the allocation of resources in the context of a public health emergency.

Key words: public health emergencies, spatiotemporal attention mechanism, spatiotemporal sequence dependence, emergency medical supplies demand forecasting

CLC Number:

C934

Xunjie Gou,Yunying Zhao,Haoyu Zhang, et al. Research on Forecasting Modelling of Emergency Medical Supplies Demand Based on Spatiotemporal Attention Mechanism[J]. Chinese Journal of Management Science, 2026, 34(7): 71-83.

Figures/Tables 11

符号	具体描述
$X = X 1, X 2, . . ., X N ∈ R N × t × d$	过去 $t$ 天内 $N$ 个地区的疫情数据
$X i = x 1 i, x 2 i, . . ., x t i ∈ R t × d$	第 $i$ 个地区在过去 $t$ 天时间范围内的疫情数据
$x j i ∈ R d$	第 $i$ 个地区在第 $j$ 天的疫情数据
$Y = Y 1, Y 2, . . ., Y N ∈ R N × τ$	未来 $τ$ 天内 $N$ 个地区的每日新增确诊病例数的真实值
$Y i = y 1 i, y 2 i, . . ., y τ i ∈ R τ$	第 $i$ 个地区在未来 $τ$ 天内的每日新增确诊病例数的真实值
$Y ̂ = Y ̂ 1, Y ̂ 2, . . ., Y ̂ N ∈ R N × τ$	未来 $τ$ 天内 $N$ 个地区的每日新增确诊病例数的预测值
$Y ̂ i = y ̂ 1 i, y ̂ 2 i, . . ., y ̂ τ i ∈ R τ$	第 $i$ 个地区在未来 $τ$ 天内的每日新增确诊病例数的预测值

References 39

[1]	Xiao C， Zhou J， Huang J， et al. C-watcher： A framework for early detection of high-risk neighborhoods ahead of COVID-19 outbreak［J/OL］. arXiv，2020， .
[2]	Kumar M， Gupta S， Kumar K， et al. Spreading of COVID-19 in India， Italy， Japan， Spain， UK， US： A prediction using ARIMA and LSTM model［J］. Digital Government： Research and Practice， 2020， 1（4）： 1-9.
[3]	Achterberg M A， Prasse B， Ma L， et al. Comparing the accuracy of several network-based COVID-19 prediction algorithms［J］. International Journal of Forecasting， 2022， 38（2）： 489-504.
[4]	Wang P， Zheng X， Li J， et al. Prediction of epidemic trends in COVID-19 with logistic model and machine learning technics［J］. Chaos， Solitons & Fractals， 2020， 139： 110058.
[5]	Wang X， Ma Y， Wang Y， et al. Traffic flow prediction via spatial temporal graph neural network［C］//Proceedings of The Web Conference 2020，Taipei，China， April 20-24， ACM， 2020： 1082-1092.
[6]	黎维，陶蔚，周星宇，等. 时空序列预测方法综述［J］. 计算机应用研究， 2020， 37（10）： 2881-2888.
	Li W， Tao W， Zhou X Y， et al. Survey of spatio-temporal sequence prediction methods［J］. Application Research of Computers， 2020， 37（10）： 2881-2888.
[7]	Lei S， Jiang F， Su W， et al. Clinical characteristics and outcomes of patients undergoing surgeries during the incubation period of COVID-19 infection［J］. EClinicalMedicine， 2020， 21： 100331.
[8]	Biswas K， Khaleque A， Sen P. Covid-19 spread： Reproduction of data and prediction using a SIR model on Euclidean network［J/OL］. arXiv， 2020，.
[9]	Mastakouri A A， Schölkopf B. Causal analysis of covid-19 spread in Germany［C］//Proceedings of the 34th International Conference on Neural Information Processing Systems， Vancouver， BC， Canada， December 6-12， ACM， 2020： 3153-3163.
[10]	Rauf H T， Lali M I U， Khan M A， et al.Retracted article： Time series forecasting of COVID-19 transmission in Asia Pacific countries using deep neural networks［J］. Personal and Ubiquitous Computing， 2023， 27（3）： 733-750.
[11]	Nguyen K L P， Chen H W， Yang C T， et al. Implementation of a respiratory disease forecasting model using LSTM for central Taiwan ［C］//Information Science and Applications， Singapore，December 19，2019， Springer， 2020： 441-450.
[12]	陈兴志，田宝单，王代文，等. 基于SEIR模型的COVID-19疫情防控效果评估和预测［J］. 应用数学和力学， 2021， 42（2）： 199-211.
	Chen X Z， Tian B D， Wang D W， et al. Evaluation and prediction of prevention and control effects of the COVID-19 epidemic based on the SEIR model［J］. Applied Mathematics and Mechanics， 2021， 42（2）： 199-211.
[13]	鲍昕，谭智一，鲍秉坤，等. 基于时空注意力机制的新冠肺炎疫情预测模型［J］. 北京航空航天大学学报， 2022， 48（8）： 1495-1504.
	Bao X， Tan Z Y， Bao B K， et al. Prediction model of COVID-19 based on spatiotemporal attention mechanism［J］. Journal of Beijing University of Aeronautics and Astronautics， 2022， 48（8）： 1495-1504.
[14]	Wang J. Spatial dynamics of an epidemic of severe acute respiratory syndrome in an urban area［J］. Bulletin of the World Health Organization， 2006， 84（12）： 965-968.
[15]	沈世敬，崔晓鸣，曹务春. 传染病预测模型的应用及研究进展［J］. 中华医院感染学杂志， 2023， 33（16）： 2550-2554.
	Shen S J， Cui X M， Cao W C. Application of prediction model for infectious diseases and progress of research［J］. Chinese Journal of Nosocomiology， 2023， 33（16）： 2550-2554.
[16]	程雪军. 新冠肺炎疫情对我国农村经济的冲击与应对［J］. 金融理论与教学， 2022（1）： 45-51.
	Cheng X J. The impact of COVID-19 on rural economy in China and the countermeasures［J］. Finance Theory and Teaching， 2022（1）： 45-51.
[17]	张晴. 关于新冠肺炎疫情的研究和预测［D］. 济南：山东大学， 2021.
	Zhang Q. Research and prediction of COVID-19［D］. Jinan： Shandong University， 2021.
[18]	Kastner S， Ungerleider L G. Mechanisms of visual attention in the human cortex［J］. Annual Review of Neuroscience， 2000， 23： 315-341.
[19]	Liang Y， Ke S， Zhang J， et al. GeoMAN： Multi-level attention networks for geo-sensory time series prediction［C］//Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence. Stockholm， Sweden， July 13-19 ， International Joint Conferences on Artificial Intelligence Organization， 2018： 3428-3434.
[20]	Mnih V， Heess N， Graves A， et al. Recurrent models of visual attention［C］//Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 2， Montreal， Canada， December 8 - 13， ACM， 2014： 2204-2212.
[21]	Bahdanau D， Cho K， Bengio Y. Neural machine translation by jointly learning to align and translate［J/OL］. arXiv， 2014， .
[22]	Vaswani A， Shazeer N， Parmar N， et al. Attention is all you need［C］//Proceedings of the 31st International Conference on Neural Information Processing Systems， Long Beach， December 4-9，ACM， 2017： 6000-6010.
[23]	Rachaniotis N， Dasaklis T K， Pappis C. Controlling infectious disease outbreaks： A deterministic allocation-scheduling model with multiple discrete resources［J］. Journal of Systems Science and Systems Engineering， 2017， 26（2）： 219-239.
[24]	Liu M， Zhang Z， Zhang D. A dynamic allocation model for medical resources in the control of influenza diffusion［J］. Journal of Systems Science and Systems Engineering， 2015， 24（3）： 276-292.
[25]	葛洪磊，刘南. 重大传染病疫情演化情境下应急物资配置决策建模分析：以新冠肺炎疫情为例［J］. 管理工程学报， 2020， 34（3）： 214-222.
	Ge H L， Liu N. Modeling of emergency materials allocation decision-making problems based on the evolution scenarios of serious infectious disease： A case of COVID-19［J］. Journal of Industrial Engineering and Engineering Management， 2020， 34（3）： 214-222.
[26]	Rawls C G， Turnquist M A. Pre-positioning of emergency supplies for disaster response［J］. Transportation Research Part B： Methodological， 2010， 44（4）： 521-534.
[27]	闫森，齐金平. 考虑需求不确定的多级应急物流设施选址研究［J］. 运筹与管理， 2022， 31（9）： 7-13.
	Yan S， Qi J P. Research on location selection of multi-level emergency logistics facilities under uncertain conditions［J］. Operations Research and Management Science， 2022， 31（9）： 7-13.
[28]	张得志，乔馨，李双艳，等. 考虑多重覆盖的应急设施多级协同布局鲁棒优化［J］. 控制与决策， 2022， 37（7）： 1853-1861.
	Zhang D Z， Qiao X， Li S Y， et al. Robust optimization of hierarchical cooperative layout of emergency facilities considering multiple coverage［J］. Control and Decision， 2022， 37（7）： 1853-1861.
[29]	Ekici A， Keskinocak P， Swann J L. Modeling influenza pandemic and planning food distribution［J］. Manufacturing & Service Operations Management， 2013， 16（1）： 11-27.
[30]	白雪. 面向重大疫情的区域应急物资需求预测与调度［J］. 物流科技， 2020， 43（8）： 87-90.
	Bai X. Demand forecast and dispatch of regional emergency supplies for major epidemic［J］. Logistics Sci-Tech， 2020， 43（8）： 87-90.
[31]	Büyüktahtakın İ E， des-Bordes E， Kıbış E Y. A new epidemics–logistics model： Insights into controlling the Ebola virus disease in West Africa［J］. European Journal of Operational Research， 2018， 265（3）： 1046-1063.
[32]	李清，苏强. 新冠肺炎疫情演化情境下应急物资需求预测研究［J］. 华中师范大学学报（自然科学版）， 2021， 55（4）： 661-670.
	Li Q， Su Q. Research on the medical allocation decisions in the context of COVID-19 epidemic evolution［J］. Journal of Central China Normal University （Natural Sciences）， 2021， 55（4）： 661-670.
[33]	Dong E， Du H， Gardner L. An interactive web-based dashboard to track COVID-19 in real time［J］. The Lancet Infectious Diseases， 2020， 20（5）： 533-534.
[34]	Gupta R， Pandey G， Chaudhary P， et al. Machine learning models for government to predict COVID-19 outbreak［J］. Digital Government： Research and Practice， 2020， 1（4）： 1-6.
[35]	Hasell J， Mathieu E， Beltekian D， et al. A cross-country database of COVID-19 testing［J］. Scientific Data， 2020， 7： 345.
[36]	Diebold F X， Mariano R S. Comparing predictive accuracy［J］. Journal of Business & Economic Statistics， 1995， 13（3）： 253-263.
[37]	新华社. 抗击新冠肺炎疫情的中国行动［EB/OL］.（2020-06-07）［2024-01-24］..
	Xinhua News Agency Fighting Covid-19 China in action［EB/OL］.（2020-06-07）［2024-01-24］..
[38]	Sutskever I， Vinyals O， Le Q V. Sequence to sequence learning with neural networks［C］//Proceedings of the 27th International Conference on Neural Information Processing Systems， Cambridge， December 5-10， 2013，ACM， 2014： 3104–3112.
[39]	Breiman L. Random forests［J］. Machine Learning， 2001， 45： 5-32.

数据项	描述
时间范围	2020/7/16-2022/4/26
地区数目	227
预测目标	每日新增确诊数
数据维度	13
数据集划分比例	8:2

指标	模型	法国	德国	印度	美国	英国	平均值
MAE	ARIMA	856.32	914.55	603.51	1946.04	304.26	924.94
	LSTM	1026.35	459.92	441.29	647.77	289.75	573.02
	GRU	1065.50	380.63	508.34	619.05	268.54	568.41
	Seq2Seq	732.27	787.64	591.37	1530.22	259.76	780.25
	Random Forest	1526.86	1477.81	752.53	2019.19	650.93	1285.46
	SpacetimeNet	467.41	348.15	96.49	738.55	347.14	399.55
RMSE	ARIMA	1178.02	1194.49	1321.86	2972.45	783.40	1490.04
	LSTM	1387.13	645.92	745.93	1457.84	473.10	941.98
	GRU	1437.16	550.58	753.49	1201.04	469.33	882.32
	Seq2Seq	1101.08	988.73	1094.56	2847.83	393.59	1285.16
	Random Forest	1965.28	1726.27	1111.74	3532.59	807.56	1828.69
	SpacetimeNet	779.48	481.00	181.85	1432.46	557.71	686.50
SMAPE	ARIMA	0.51	0.64	1.20	1.08	0.77	0.84
	LSTM	0.85	0.36	0.64	0.31	0.76	0.59
	GRU	0.89	0.32	0.66	0.31	0.74	0.58
	Seq2Seq	0.49	0.63	1.08	1.05	0.65	0.78
	Random Forest	1.60	1.73	1.38	0.95	1.48	1.43
	SpacetimeNet	0.35	0.31	0.15	0.43	0.73	0.40

模型		ARIMA	LSTM	GRU	Seq2Seq	Random Forest
法国	DM	-4.97	-6.63	-6.34	-5.29	-7.33
法国	p_value	2.44E-06	1.34E-09	5.09E-09	6.20E-07	4.25E-11
德国	DM	-3.66	-2.88	-2.77	-3.84	-7.61
德国	p_value	3.78E-04	4.76E-03	6.52E-03	2.03E-04	1.04E-11
印度	DM	-6.26	-4.8	-5.07	-5.39	-5.55
印度	p_value	7.53E-09	4.91E-06	1.58E-06	4.05E-07	1.95E-07
美国	DM	-4.2	-3.24	-2.21	-4.49	-4.61
美国	p_value	5.32E-05	1.56E-03	2.89E-02	1.74E-05	1.09E-05
英国	DM	-1.31	-3.83	-3.39	-3.35	-5.96
英国	p_value	1.60E-02	2.08E-04	9.57E-04	1.08E-03	3.15E-08

指标	模型	法国	德国	印度	美国	英国	平均
MAE	SpacetimeNet	467.41	348.15	96.49	738.55	347.14	399.55
	SpacetimeNet_NS	737.19	501.77	386.24	1093.27	354.72	614.64
	SpacetimeNet_NT	685.64	587.20	109.31	974.21	351.57	541.59
	SpacetimeNet_NA	1277.67	639.08	442.56	2300.85	652.09	1062.45
	SpacetimeNet_NM	1623.55	1541.13	738.31	2342.49	735.53	1396.20
RMSE	SpacetimeNet	779.48	481.00	181.85	1432.46	557.71	686.50
	SpacetimeNet_NS	1024.04	715.27	705.32	1945.12	543.25	986.60
	SpacetimeNet_NT	1033.90	818.59	186.71	1836.48	554.35	886.01
	SpacetimeNet_NA	1725.17	855.31	860.77	3431.05	1007.99	1576.06
	SpacetimeNet_NM	2044.24	1782.50	1279.31	3894.42	904.51	1981.00
SMAPE	SpacetimeNet	0.35	0.31	0.15	0.43	0.73	0.40
	SpacetimeNet_NS	0.58	0.38	0.44	0.53	0.76	0.53
	SpacetimeNet_NT	0.47	0.43	0.31	0.52	0.75	0.49
	SpacetimeNet_NA	1.07	0.48	0.61	1.09	0.84	0.82
	SpacetimeNet_NM	1.98	1.97	1.74	1.94	1.97	1.92

模型		Spacetimenet_NS	Spacetimenet_NT	Spacetimenet_NA	Spacetimenet_NM
法国	DM	-5.43	-4.14	-6.62	-7.51
法国	p_value	3.43E-07	6.80E-05	1.34E-09	1.65E-11
德国	DM	-3.30	-1.09	-1.53	-8.83
德国	p_value	1.27E-03	2.76E-02	1.28E-02	1.96E-14
印度	DM	-5.02	-0.54	-4.86	-5.64
印度	p_value	1.99E-06	5.87E-01	3.94E-06	1.32E-07
美国	DM	-2.73	-1.57	-5.09	-5.22
美国	p_value	7.20E-03	1.18E-03	1.49E-06	8.35E-07
英国	DM	-4.60	-3.20	-2.40	-3.93
英国	p_value	1.11E-05	1.78E-03	1.79E-02	1.48E-04

Research on Forecasting Modelling of Emergency Medical Supplies Demand Based on Spatiotemporal Attention Mechanism

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 11

References 39

Related Articles 5

Metrics

Comments

Recommended 0

[1]	Yiying Li, Boqun Wang. Precision Intervention and Household Consumption during Public Health Emergencies: A Two-way Fixed Effects Model Analysis [J]. Chinese Journal of Management Science, 2025, 33(5): 225-235.
[2]	Fangju Jia,Kun Zhou,Lianshui Li. Stochastic Evolutionary Decision Analysis of Collaborative Prevention and Control Strategies for Public Health Emergencies [J]. Chinese Journal of Management Science, 2024, 32(3): 237-247.
[3]	Guirong Zhang,Zhixiang Dong,Ting Xia. Study on the Grid-based Early Warning Model for Online Rumors of Public Health Emergencies [J]. Chinese Journal of Management Science, 2024, 32(2): 298-306.
[4]	ZHU Hong-miao, YAN Xin, QI Jia-yin, JIN Zhen. Research on Communication Model and Intervention Strategy of COVID-19 Vaccination Consciousness Among Public [J]. Chinese Journal of Management Science, 2023, 31(8): 269-277.
[5]	WEN Feng-hua, SHEN Ti-yan, XING Jiang-bo, KOU Chen-huan. Urban Evidence-based Governance Mechanism of Public Health Emergencies [J]. Chinese Journal of Management Science, 2023, 31(1): 206-215.