Anticipatory Risk Analytics for Global Response on the Containment of COVID-19

Abstract

The Prediction & Control Problem: Typically, prediction is based upon historical data where plenty of data is available over extended time duration within relatively "static" linear and normal contexts. Such relatively deterministic, and, [statistically] linear and normal contexts are suitable for typical AI-ML-DL driven analytics and data science driven prediction based on history. The Prediction Problem occurs wherein either past data is unavailable or is sparse as in case of COVID 19 wherein future prediction is based upon negligible to sparse [but increasingly cumulative] context-specific data in real time while the specific contexts are dynamically evolving. Prediction is typically used for Control as in the context of "flattening the curve" [which is a function of both minimizing the "spread" of the COVID 19 risk such as by using 'social distancing' while maximizing the "capacity" to mitigate the COVID 19 risk such as by increasing hospital bed capacity] in context of each "hot spot" [Different 'hot spots' may be characterized by differences in severity and intensity of the outbreak risk given context-specific determinants such as density and connectedness that may determine the rate and speed of spread of such risk.]

Keywords: COVID-19, Coronavirus, Healthcare Prediction, Control, Models, Modeling, Model Risk Management, Catastrophic Risk