New MIT ML Model Shows That Easing Restrictions Will Spike Covid-19 Cases

Based on epidemiological equations established around outbreaks and data-driven neural network-based inference, MIT has developed a new machine learning (ML) model around the spread of Covid-19.

The study had narrowed down its focus to four areas - Wuhan, Italy, South Korea and the US. With publicly available data on these locales, the researchers compared the role that quarantine measures have played in controlling the effective reproduction number (RO) of the virus.

“By relaxing quarantine measures too soon, we have predicted that the consequences would be far more catastrophic,” said model developer and MIT mechanical engineering professor, George Barbastathis.

ML Model Demonstrates Risk Of Easing Restrictions

This study was conducted on the backdrop of countries like Singapore which had lifted quarantine measures after cases levelled off, and thereafter, experienced a resurgence in Covid-19 cases. The MIT researchers developed the model based on data available on Covid-19 alone. This is unlike most others, who have used SARS or MERS information to inform their charting of the outbreak’s progress.

Combining this information with a neural network-based estimation of the number of infected individuals under quarantine, has allowed them to go beyond existing models in terms of accurately modelling and predicting the effect of isolation measures.

It has also helped them understand what could happen if those measures are withdrawn. The model proves accurate when trained on this data, and indicates that any immediate relaxation of quarantine measures would lead to an “exponential explosion” in the number of infections.