Das Monidipa, Ghosh Akash, Ghosh Soumya K
Machine Intelligence Unit (MIU), Indian Statistical Institute (ISI), Kolkata, India.
Department of Computer Science and Engineering, Jadavpur University (JU), Kolkata, India.
SN Comput Sci. 2021;2(6):452. doi: 10.1007/s42979-021-00845-9. Epub 2021 Sep 9.
COVID-19, a life-threatening infection by novel coronavirus, has broken out as a pandemic since December 2019. Eventually, with the aim of helping the World Health Organization and other health regulators to combat COVID-19, significant research effort has been exerted during last several months to analyze how the various factors, especially the climatic aspects, impact on the spread of this infection. However, due to insufficient test and lack of data transparency, these research findings, at times, are found to be inconsistent as well as conflicting. In our work, we aim to employ a semantics-driven probabilistic framework for analyzing the causal influence as well as the impact of climate variability on the COVID-19 outbreak. The idea here is to tackle the data inadequacy and uncertainty issues using probabilistic graphical analysis along with embedded technology of incorporating semantics from climatological domain. Furthermore, the theoretical guidance from epidemiological model additionally helps the framework to better capture the pandemic characteristics. More significantly, we further enhance the impact analysis framework with an auxiliary module of measuring semantic relatedness on regional basis, so as to realistically account for the existence of multiple climate types within a single spatial region. This added notion of regional semantic relatedness further helps us to attain improved probabilistic analysis for modeling the climatological impact on this disease outbreak. Experimentation with COVID-19 datasets over 15 states (or provinces) belonging to varying climate regions in India, demonstrates the effectiveness of our semantically-enhanced theory-guided data-driven approach. It is worth noting that our proposed framework and the relevant semantic analyses are generic enough for intelligent as well as explainable impact analysis in many other application domains, by introducing minimal augmentation.
新型冠状病毒引发的危及生命的感染病COVID-19自2019年12月以来已爆发成为全球大流行病。最终，为了帮助世界卫生组织和其他卫生监管机构抗击COVID-19，在过去几个月里人们付出了巨大的研究努力，以分析各种因素，特别是气候因素，如何影响这种感染病的传播。然而，由于检测不足和数据缺乏透明度，这些研究结果有时被发现是不一致且相互矛盾的。在我们的工作中，我们旨在采用一个语义驱动的概率框架来分析气候变率对COVID-19爆发的因果影响及其作用。这里的想法是利用概率图形分析以及纳入气候学领域语义的嵌入式技术来解决数据不足和不确定性问题。此外，流行病学模型的理论指导进一步帮助该框架更好地捕捉大流行特征。更重要的是，我们通过一个在区域基础上测量语义相关性的辅助模块进一步增强了影响分析框架，以便切实考虑单个空间区域内多种气候类型的存在。这种区域语义相关性的附加概念进一步帮助我们获得改进的概率分析，以模拟气候对这种疾病爆发的影响。对印度不同气候区域的15个邦（或省）的COVID-19数据集进行的实验，证明了我们语义增强的理论指导数据驱动方法的有效性。值得注意的是，我们提出的框架和相关语义分析具有足够的通用性，通过引入最少的扩充，可用于许多其他应用领域的智能且可解释的影响分析。
