论文标题
通过机器学习分析印度的Covid-19案件：干预研究
Analysis of COVID-19 cases in India through Machine Learning: A Study of Intervention
论文作者
论文摘要
为了应对2019年冠状病毒疾病（Covid-19）的大流行，全世界都有疫苗接种，血浆疗法，牛群免疫和流行病学干预措施，这是少数可能的选择。 Covid-19-19疫苗开发正在进行中，开发疫苗可能需要大量时间，开发后，将需要时间才能接种整个人群，并且血浆疗法有一些局限性。畜群的免疫力可能是为小国与COVID战斗的合理选择。但是，对于像印度这样人口庞大的国家而言，牛群的免疫力不是一个合理的选择，因为要获得大约67％的人口的群体豁免权必须从COVID-19-19的感染中恢复，这将给该国的医疗体系带来额外的负担，并会导致人类生活巨大的损失。因此，流行病学干预措施（完全锁定，部分锁定，隔离，隔离，社会距离等）是印度的一些合适策略，以减慢19009年的速度，直到疫苗开发。在这项工作中，我们提出了带有干预措施的SIR模型，该模型将流行病学干预措施纳入经典的SIR模型中。为了建模干预措施的效果，我们将\ r {ho}作为干预参数。 \ r {ho}是一个累积数量，涵盖了所有类型的干预措施。我们还讨论了监督的机器学习方法，以估计SIR模型的传输速率（\ b {eta}），并从印度和印度某些州的COVID-19数据的患病率中进行干预。为了验证我们的模型，我们提出了实际和模型预测的COVID-19案例的比较。使用我们的模型，我们还向整个印度和印度某些州提供了预测的活跃和回收的Covid-19案例，直到2020年9月30日，还估计预测案件的置信区间为95％和99％。
To combat the coronavirus disease 2019 (COVID-19) pandemic, the world has vaccination, plasma therapy, herd immunity, and epidemiological interventions as few possible options. The COVID-19 vaccine development is underway and it may take a significant amount of time to develop the vaccine and after development, it will take time to vaccinate the entire population, and plasma therapy has some limitations. Herd immunity can be a plausible option to fight COVID-19 for small countries. But for a country with huge population like India, herd immunity is not a plausible option, because to acquire herd immunity approximately 67% of the population has to be recovered from COVID-19 infection, which will put an extra burden on medical system of the country and will result in a huge loss of human life. Thus epidemiological interventions (complete lockdown, partial lockdown, quarantine, isolation, social distancing, etc.) are some suitable strategies in India to slow down the COVID-19 spread until the vaccine development. In this work, we have suggested the SIR model with intervention, which incorporates the epidemiological interventions in the classical SIR model. To model the effect of the interventions, we have introduced \r{ho} as the intervention parameter. \r{ho} is a cumulative quantity which covers all type of intervention. We have also discussed the supervised machine learning approach to estimate the transmission rate (\b{eta}) for the SIR model with intervention from the prevalence of COVID-19 data in India and some states of India. To validate our model, we present a comparison between the actual and model-predicted number of COVID-19 cases. Using our model, we also present predicted numbers of active and recovered COVID-19 cases till Sept 30, 2020, for entire India and some states of India and also estimate the 95% and 99% confidence interval for the predicted cases.
