A Heavily Trained Time-Dependent SIRD Model for Local Covid-19 Data in Italy

Abstract

We present a time-dependent SIRD model for the spread of COVID-19 infection at a provincial (i.e. EU NUTS-3) level in Italy, using official data from the Italian Ministry of Health, integrated with data extracted from daily official press conferences of regional authorities and from local newspaper websites. This integration concerns COVID-19 death data which are not available at NUTS-3 level from open official data channels. The model is trained for improved forecasting performance with similarity techniques putting together data from time series most similar to that for which the forecast is performed.

Provincial Level Time Series Analysis : Community Mobility and COVID-19 Daily Cases in Turin, Italy (Preprint)

“…The standard Susceptible-Infected-Recovery-Death (SIRD) model [10] is a classic epidemiological model that was utilized in this study for analysing the COVID-19 infection in Turin during the first wave. This compartmental model divides the whole population into four different groups such as susceptible (S), currently infected (I), recovered (R) and deaths (D) for all times t where N is the size of the population [11] and it is denoted as N = S + I + R + D. The parameters governing this model are transmission rate (β), recovery rate( γ), the mortality rate (δ). Interaction of a susceptible individual with a SARS-CoV-2 positive patient makes the infection rate of the susceptible individual at a rate of β S I/N [12].…”