Thomas is a writer at TNW. He covers the full spectrum of European tech, with a particular focus on deeptech, startups, and government polic Thomas is a writer at TNW. He covers the full spectrum of European tech, with a particular focus on deeptech, startups, and government policy.
An AI has predicted which areas of the UK will suffer spikes in coronavirus cases next by analyzing social media posts about COVID-19.
These posts suggest that outbreaks will soon hit Manchester, Liverpool, Oxfordshire, Leeds, Northamptonshire, and Luton.
The system was created by Dataminr, a risk detection tech firm with a track record of detecting where the virus will spread to next.
It previously predicted outbreaks in London, Hertfordshire, Essex, and Kent, between 7-13 days before these areas had spikes in coronavirus cases. In the US, Dataminr forecast outbreaks in 14 different states. Seven days after the company published its predictions, all states had been hit hard by the pandemic seven days later.
[Read: MIT researchers use AI to turn the coronavirus into a haunting melody]
The same method has now spotted similar pre-outbreak patterns in six regions of the UK.
Predicting the pandemic
Dataminr’s AI divided first-hand social media posts about the coronavirus into different geo-localized hotspots.
The system doesn’t measure aggregate social media chatter or keyword mentions related to the pandemic. Instead, it finds unique posts from individuals indicating they have coronavirus symptoms, have tested positive or been exposed to the virus. It also includes first-hand accounts of confirmed cases from their relatives, friends, and coworkers.
This approach aims to cut out gossip and misleading COVID-19 references to create a ground truth picture of how the outbreak is progressing.
The findings will be hard to hear for the regions where spikes are expected. But they at least give governments, companies, and citizens new insights that can help them plan their responses.
Get the TNW newsletter
Get the most important tech news in your inbox each week.