[Paper Review] Day Level Forecasting for Coronavirus Disease (COVID-19) Spread: Analysis, Modeling and Recommendations

Haytham H. Elmousalami, Aboul Ella Hassanien|arXiv (Cornell University)|Mar 15, 2020

COVID-19 epidemiological studies11 references68 citations

TL;DR

This paper compares day-level forecasting models for COVID-19 spread using time-series and mathematical formulations and discusses the impact of social distancing measures on case growth.

ABSTRACT

In mid of March 2020, Coronaviruses such as COVID-19 is declared as an international epidemic. More than 125000 confirmed cases and 4,607 death cases have been recorded around more than 118 countries. Unfortunately, a coronavirus vaccine is expected to take at least 18 months if it works at all. Moreover, COVID -19 epidemics can mutate into a more aggressive form. Day level information about the COVID -19 spread is crucial to measure the behavior of this new virus globally. Therefore, this study presents a comparison of day level forecasting models on COVID-19 affected cases using time series models and mathematical formulation. The forecasting models and data strongly suggest that the number of coronavirus cases grows exponentially in countries that do not mandate quarantines, restrictions on travel and public gatherings, and closing of schools, universities, and workplaces (Social Distancing).

Motivation & Objective

• Motivate the importance of day-level data to understand COVID-19 spread globally in early 2020.
• Compare forecasting models at daily resolution using time-series and mathematical formulations.
• Provide recommendations based on model analyses and observed growth patterns under different policy responses.

Proposed method

• Compare day-level forecasting models on COVID-19 affected cases using time-series models and mathematical formulations.
• Use exponential-growth considerations to analyze spread dynamics.
• Draw recommendations based on model analyses and observed policy impacts such as social distancing.

Experimental results

Research questions

• RQ1 How do day-level forecasting models perform on COVID-19 case data across countries?
• RQ2 What is the role of interventions (quarantines, travel restrictions, school/workplace closures) in shaping day-level growth patterns?
• RQ3 Can time-series and mathematical formulation approaches capture the exponential growth observed during early COVID-19 spread?

Key findings

• The number of coronavirus cases grows exponentially in countries that do not mandate quarantines, travel restrictions, and closures of schools, universities, and workplaces.
• The study presents a comparison of day-level forecasting models using time-series and mathematical formulations.
• Forecasting models and data suggest policy measures influence the growth dynamics of COVID-19 spread.