[Paper Review] Forecasting COVID 19 growth in India using Susceptible-Infected-Recovered (S.I.R) model

Jay Naresh Dhanwant, V. Ramanathan|arXiv (Cornell University)|Apr 1, 2020

COVID-19 epidemiological studiesMathematics4 references36 citations

TL;DR

The paper uses a Susceptible-Infected-Recovered (SIR) model with a learnable social-contact parameter eta to forecast COVID-19 growth in India and assess the effectiveness of social distancing measures. It concludes that the observed social distancing in India at the time was insufficient to curb growth.

ABSTRACT

This work covers the analysis of the COVID 19 spread in different countries and dealing the main feature of COVID 19 growth, which is the spread due to the social-contact structure, which is governed by the parameter \b{eta}. The dependency of this parameter \b{eta} on the transmission level in society gives a sense of the effectiveness of the measures taken for social distancing. A separate algorithm is hardcoded in python using Scipy which learns the social-contact structure and gives a suitable value for \b{eta}, which has a major impact on the outcome of the result. Forecasting for the epidemic spread in India was done, and it was found that the strictness at which social distancing in India is done, is insufficient for the growth of COVID 19.

Motivation & Objective

• Analyze COVID-19 spread in India using an SIR framework.
• Model how social-contact structure (eta) affects transmission.
• Develop a Python/Scipy-based algorithm to learn eta from data.
• Forecast epidemic growth under Indian social-distancing scenarios.

Proposed method

• Adopt an SIR model where the transmission parameter depends on social-contact structure (eta).
• Hardcode a Python/Scipy-based algorithm to learn eta from data.
• Use the learned eta to forecast epidemic growth in India.
• Evaluate the impact of social-distancing strictness on growth outcomes.

Experimental results

Research questions

• RQ1 How does the social-contact structure parameter eta influence COVID-19 transmission in an SIR framework?
• RQ2 Can a dedicated algorithm infer eta from empirical data to improve forecasts?
• RQ3 What do the forecasts imply about the sufficiency of social-distancing measures in India during the study period.

Key findings

• A learnable eta parameter captures social-contact structure and affects model outcomes.
• The algorithm estimates eta to reflect transmission level and social-distancing effectiveness.
• Forecasts suggest that the social-distancing strictness in India at the time was insufficient for controlling growth.