Data-Driven Models for studying the Dynamics of the COVID-19 Pandemics
This paper seeks to study the evolution of the COVID-19 pandemic based on daily published data from Worldometer website, using a time-dependent SIR model. Our findings indicate that this model fits well such data, for different chosen periods and different regions. This well-known model, consisting of three disjoint compartments, susceptible , infected , and removed , depends in our case on two time dependent parameters, the infection rate $β(t)$ and the removal rate $ρ(t)$. After deriving the model, we prove the local exponential behavior of the number of infected people, be it growth or decay. Furthermore, we extract a time dependent replacement factor $σ_s(t) ={β(t)}s(t)/{ρ(t) }$, where $s(t)$ is the ratio of susceptible people at time $t$. In addition, $i(t)$ and $r(t)$ are respectively the ratios of infected and removed people, based on a population of size $N$, usually assumed to be constant. Besides these theoretical results, the report provides simulations on the daily data obtained for Germany, Italy, and the entire World, as collected from Worldometer over the period stretching from April 2020 to June 2022. The computational model consists of the estimation of $β(t)$, $ρ(t)$ and $s(t)$ based on the time-dependent SIR model. The validation of our approach is demonstrated by comparing the profiles of the collected $i(t), r(t)$ data and those obtained from the SIR model with the approximated parameters. We also consider matching the data with a constant-coefficient SIR model, which seems to be working only for short periods. Thus, such model helps understanding and predicting the evolution of the pandemics for short periods of time where no radical change occurs.