Deconfining? Reconfigure? Open schools? To restore economy? These are all complex questions to which politicians must provide answers in the face of the Covid-19 pandemic.
An international team of scientists has developed a model algorithm that could predict the movements of the epidemic about two weeks before they occur. So early enough for leaders to take the appropriate action, according to The New York Times.
A team of Harvard scientists, led by Mauricio Santillana and Nicole Kogan, presented an algorithm that registers the danger at least 14 days before the number of cases begins to increase. The system uses, among other things, real-time monitoring of Twitter, Google searches and mobility data from smartphones.
“What we are doing here is observing, without making assumptions. Our methods are therefore more sensitive to immediate changes in behavior, which we can quickly integrate into the algorithm. ”, explains Mauricio Santillana, director of the artificial intelligence laboratory at the Boston Children’s Hospital and assistant professor of pediatrics and epidemiology at Harvard.
The study shows “That new generation alternative data sources can provide early signals of the increasing prevalence of Covid-19”said Lauren Ancel Meyers, biologist and statistician at the University of Texas, Austin. “Particularly if the number of confirmed cases is delayed by seeking treatment or obtaining test results.”
In New York, for example, the number of Twitter messages linked to Covid-19 jumped more than a week before the number of cases exploded in mid-March.
… will wait
Despite all its appeal, data analysis cannot anticipate sudden changes in the behavior of the masses. No algorithm could have predicted the national protests that followed the murder of George Floyd, for example.
Social networks and search engines may also become less sensitive over time. The more people are familiar with a pathogen, the less they will search for related keywords.
“These are extremely valuable data we have”said Shweta Bansal, a biologist at Georgetown University. “But I don’t want to apply [l’algorithme] actually right now. It’s too early, the risks involved are too great. We need to see these models verified and validated over time. ”
“What we looked at was what we think are the best data feeds available today”, notes Dr. Santillana. “We can’t wait to see what Amazon or Netflix can give us.” U.S. too…