Using Data Science to monitor the pandemic with a single number: the Synthetic COVID Index

TL;DR

The paper introduces the Synthetic COVID Index, an innovative single-number metric derived from unsupervised machine learning, to effectively summarize and monitor the pandemic's severity using complex data.

Contribution

It presents a novel ensemble unsupervised learning approach to estimate a latent variable representing pandemic strength, aiding policymakers with simplified data interpretation.

Findings

The index effectively summarizes the pandemic situation in Italy.

It captures the pandemic's strength despite measurement errors.

The method provides a concise and interpretable indicator.

Abstract

Rapid and affordable methods of summarizing the multitude of data relating to the pandemic can be useful to health authorities and policy makers who are dealing with the COVID-19 pandemic at various levels in the territories affected by SARSCoV-2. This is the goal of the Synthetic COVID Index, an index based on an ensemble of Unsupervised Machine Learning techniques which focuses on the identification of a latent variable present in data that contains measurement errors. This estimated latent variable can be interpreted as "the strength of the pandemic". An application to the Italian case shows how the index is able to provide a concise representation of the situation.