A critical analysis of the open data provided by the Italian Civil Protection Centre during phase 1 of Covid-19 epidemic—the so-called Italian lockdown—is herein proposed in relation to four of the most affected Italian regions, namely Lombardy, Reggio Emilia, Valle d’Aosta, and Veneto. A possible bias in the data induced by the extent in the use of medical swabs is found in relation to Valle d’Aosta and Veneto. Observed data are then interpreted using a Susceptible-Infectious-Recovered (SIR) epidemiological model enhanced with asymptomatic (infected and recovered) compartments, including lockdown effects through time-dependent model parameters. The initial number of susceptible individuals for each region is also considered as a parameter to be identified. The issue of parameters identification is herein addressed by a robust machine learning approach based on particle swarm optimization. Model predictions provide relevant information for policymakers in terms of the effect of lockdown measures in the different regions. The number of susceptible individuals involved in the epidemic, important for a safe release of lockdown during the next phases, is predicted to be around 10% of the population for Lombardy, 16% for Reggio Emilia, 18% for Veneto, and 40% for Valle d’Aosta.
An Analysis of the Italian Lockdown in Retrospective Using Particle Swarm Optimization in Machine Learning Applied to an Epidemiological Model
Paggi M.
2020-01-01
Abstract
A critical analysis of the open data provided by the Italian Civil Protection Centre during phase 1 of Covid-19 epidemic—the so-called Italian lockdown—is herein proposed in relation to four of the most affected Italian regions, namely Lombardy, Reggio Emilia, Valle d’Aosta, and Veneto. A possible bias in the data induced by the extent in the use of medical swabs is found in relation to Valle d’Aosta and Veneto. Observed data are then interpreted using a Susceptible-Infectious-Recovered (SIR) epidemiological model enhanced with asymptomatic (infected and recovered) compartments, including lockdown effects through time-dependent model parameters. The initial number of susceptible individuals for each region is also considered as a parameter to be identified. The issue of parameters identification is herein addressed by a robust machine learning approach based on particle swarm optimization. Model predictions provide relevant information for policymakers in terms of the effect of lockdown measures in the different regions. The number of susceptible individuals involved in the epidemic, important for a safe release of lockdown during the next phases, is predicted to be around 10% of the population for Lombardy, 16% for Reggio Emilia, 18% for Veneto, and 40% for Valle d’Aosta.File | Dimensione | Formato | |
---|---|---|---|
physics-02-00020-v2.pdf
accesso aperto
Tipologia:
Versione Editoriale (PDF)
Licenza:
Creative commons
Dimensione
2.89 MB
Formato
Adobe PDF
|
2.89 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.