Real-time forecasts and risk assessment of novel coronavirus (COVID-19) cases: A data-driven analysis

Tanujit Chakraborty; Indrajit Ghosh

The second problem is connected with the global concern of health and mortality due to the significant COVID-19 outbreaks. Mortality is crudely estimated using a statistic, the case fatality rate (CFR), which divides the number of known deaths by the total number of identified cases [18; 5; 30] . During the current phase of this global pandemic, it is criti-cally important to obtain reliable estimates of the overall CFR. The estimates of CFR are highly dependent on several country-specific demographic parameters and various disease characteristics. A key differentiation among the CFR of different countries can be found by determining an exhaustive list of causal variables that significantly affect CFR. In this work, we put an effort to identify critical parameters that may help to assess the risk (in terms of CFR) using an optimal regression tree model [7] . The regression tree has a built-in variable selection mechanism from high dimensional variable space and can model arbitrary decision boundaries. The regression tree combines case estimates, epidemiological characteristics of the disease, and heath-care facilities to assess the risks of major outbreaks for profoundly affected countries. Such assessments will help to anticipate the expected morbidity and mortality due to COVID-19 and provide some critical information for the planning of health care systems in various countries facing this epidemic.