An Improved SEIR Dynamics Model for Actual Infection Scale Estimation of COVID-19

Abstract

It is crucial to capture the actual infection scale of communicable diseases. However, the official case numbers cannot equal to actual infection scale in society, because a large number of asymptomatic individuals are not recognized. To deal with this challenge, this paper takes COVID-19 as the object, and develops an improved SEIR dynamics model to estimate its actual infection scale. Generalized to the circumstances in this work, we improve the classical SEIR model by considering three implicit factors: self-recovered asymptomatic individuals, recovered individuals, and deceased individuals. The dynamics process inside the improved model is expressed using mathematical formulas, and the parameter estimation scheme is given accordingly. To evaluate the estimation effect of the proposal, we employ pandemic data from 10 representative countries to build experimental scenario. The results obtained through model fitting demonstrate that the estimated actual infection scale is approximately 10–30 times higher than the reported average “newly confirmed cases”. Furthermore, our findings reveal a noteworthy negative correlation between the transmission coefficient and vaccination rate, confirming the beneficial role of vaccination in mitigating COVID-19 spreading.