Feroze Navid
Department of Statistics, The University of Azad Jammu and Kashmir, Muzffarabad, Pakistan.
Infect Dis Model. 2021;6:343-350. doi: 10.1016/j.idm.2021.01.005. Epub 2021 Jan 22.
The short term forecasts regarding different parameters of the COVID-19 are very important to make informed decisions. However, majority of the earlier contributions have used classical time series models, such as auto regressive integrated moving average (ARIMA) models, to obtain the said forecasts for Iran and its neighbors. In addition, the impacts of lifting the lockdowns in the said countries have not been studied. The aim of this paper is to propose more flexible Bayesian structural time series (BSTS) models for forecasting the future trends of the COVID-19 in Iran and its neighbors, and to compare the predictive power of the BSTS models with frequently used ARIMA models. The paper also aims to investigate the casual impacts of lifting the lockdown in the targeted countries using proposed models.
We have proposed BSTS models to forecast the patterns of this pandemic in Iran and its neighbors. The predictive power of the proposed models has been compared with ARIMA models using different forecast accuracy criteria. We have also studied the causal impacts of resuming commercial/social activities in these countries using intervention analysis under BSTS models. The forecasts for next thirty days were obtained by using the data from March 16 to July 22, 2020. These data have been obtained from Our World in Data and Humanitarian Data Exchange (HDX). All the numerical results have been obtained using R software.
Different measures of forecast accuracy advocated that forecasts under BSTS models were better than those under ARIMA models. Our forecasts suggested that the active numbers of cases are expected to decrease in Iran and its neighbors, except Afghanistan. However, the death toll is expected to increase at more pace in majority of these countries. The resuming of commercial/social activities in these countries has accelerated the surges in number of positive cases.
The serious efforts would be needed to make sure that these expected figures regarding active number of cases come true. Iran and its neighbors need to improve their extensive healthcare infrastructure to cut down the higher expected death toll. Finally, these countries should develop and implement the strict SOPs for the commercial activities in order to prevent the expected second wave of the pandemic.
对2019冠状病毒病不同参数的短期预测对于做出明智决策非常重要。然而，大多数早期研究都使用经典时间序列模型，如自回归积分滑动平均（ARIMA）模型，来对伊朗及其邻国进行上述预测。此外，上述国家解除封锁的影响尚未得到研究。本文的目的是提出更灵活的贝叶斯结构时间序列（BSTS）模型，以预测伊朗及其邻国2019冠状病毒病的未来趋势，并将BSTS模型的预测能力与常用的ARIMA模型进行比较。本文还旨在使用所提出的模型研究解除目标国家封锁的偶然影响。
我们提出了BSTS模型来预测伊朗及其邻国这种大流行病的模式。使用不同的预测准确性标准，将所提出模型的预测能力与ARIMA模型进行了比较。我们还在BSTS模型下使用干预分析研究了这些国家恢复商业/社会活动的因果影响。通过使用2020年3月16日至7月22日的数据获得了未来30天的预测。这些数据来自“Our World in Data”和人道主义数据交换（HDX）。所有数值结果均使用R软件获得。
不同的预测准确性度量表明，BSTS模型下得到的预测优于ARIMA模型下得到的预测。我们的预测表明，除阿富汗外，伊朗及其邻国的活跃病例数预计将减少。然而，这些国家中的大多数国家的死亡人数预计将以更快的速度增加。这些国家恢复商业/社会活动加速了阳性病例数激增。
需要做出认真努力，以确保这些关于活跃病例数的预期数字能够实现。伊朗及其邻国需要改善其广泛的医疗基础设施，以降低预期的更高死亡人数。最后，这些国家应制定并实施严格的商业活动标准作业程序，以防止预期的第二波疫情。
