Press Releases

AI and Social Media Forecast Bird Flu Outbreaks 14 Days Early, New Study Finds

TORONTO, CANADA — Researchers at the University of Toronto’s Artificial Intelligence and Mathematical Modeling Lab (AIMMLab) have developed a pioneering deep-learning system capable of predicting highly pathogenic avian influenza (HPAI) outbreaks in Canada up to 14 days in advance. By integrating unconventional data sources such as Facebook activity, Google searches, and weather patterns, this framework provides a powerful early-warning tool for protecting both public health and the global poultry economy.

Published in the Journal of the Royal Society Interface, the study comes at a critical time as the H5N1 subtype continues to threaten global health and food security. In humans, H5N1 carries a mortality rate of approximately 60%, while economically it has resulted in billions of dollars in losses and major disruptions to poultry supply chains worldwide.

Predicting the Unpredictable

Traditional disease surveillance systems often rely on laboratory confirmations that may take weeks. In contrast, AIMMLab’s AI-driven pipeline operates in near real time using a deep-learning architecture known as a Gated Recurrent Unit (GRU), integrating seven complementary data streams:

• Social & Web Signals: Facebook, Reddit, Google Trends, and Google News activity.
• Environmental Indicators: Minimum temperature, UV index, and carbon monoxide levels.
• Historical Surveillance: Past HPAI outbreak data.

“Our findings show that combining web-based behavioral signals with environmental data substantially enhances outbreak forecasting, enabling faster and more informed public-health interventions,” the authors report.

The Power of Digital Signals

While historical case data remains the strongest predictor, Facebook activity and minimum temperature emerged as key early indicators. The model demonstrated exceptional predictive performance at both national and regional scales, particularly in Central Canada.

This proactive forecasting approach empowers veterinary authorities and poultry producers to implement targeted biosecurity measures such as movement restrictions, preemptive vaccinations, and enhanced surveillance—reducing mass culling and minimizing trade disruptions.

A Modular Future for Digital Epidemiology

Designed to be modular and scalable, this system can be adapted for other infectious diseases and expanded to new regions globally. A real-time dashboard has also been launched to visualize predictions and support data-driven decision-making for policymakers.

About the AIMMLab

The Artificial Intelligence and Mathematical Modeling Lab (AIMMLab) at the University of Toronto, led by Professor Jude Dzevela Kong, focuses on leveraging advanced AI techniques to tackle global health challenges and strengthen infectious disease surveillance systems.