A Physics-Based Data-Driven Approach for Finite Time Estimation of Pandemic Growth

TL;DR

This paper introduces a physics-based, data-driven model combining conservation laws, quadratic programming, and neural networks to predict COVID-19 pandemic growth in the US over a finite time horizon.

Contribution

It presents a novel integration of physics-based modeling with machine learning techniques for pandemic growth prediction.

Findings

Accurately predicts COVID-19 case growth over finite time.

Validates model with US COVID-19 data from March 2020 to October 2021.

Demonstrates effectiveness of combining physics laws with neural networks.

Abstract

COVID-19 is a global health crisis that has had unprecedented, widespread impact on households across the United States and has been declared a global pandemic on March 11, 2020 by World Health Organization (WHO) [1]. According to Centers for Disease Control and Prevention (CDC) [2], the spread of COVID-19 occurs through person-to-person transmission i.e. close contact with infected people through contaminated surfaces and respiratory fluids carrying infectious virus. This paper presents a data-driven physics-based approach to analyze and predict the rapid growth and spread dynamics of the pandemic. Temporal and Spatial conservation laws are used to model the evolution of the COVID-19 pandemic. We integrate quadratic programming and neural networks to learn the parameters and estimate the pandemic growth. The proposed prediction model is validated through finite time estimation of the pandemic growth using the total number of cases, deaths and recoveries in the United States recorded from March 12, 2020 until October 1, 2021 [3].