By combining anonymized social mobility data from Google with social media signals, we will predict COVID-19 prevalence across the US and Europe. COVID-19 spreads through human social networks, making mobility data essential for predicting emerging clusters of the disease. At the same time, the speed and ease of sharing information via social media makes the latter an "early warning" system for emerging phenomena like pandemics. We will use our patented social media mapping system with deep learning models integrating social media, mobility, and epidemiological data to predict COVID-19 prevalence in geographical regions over time. Our research shows deep learning models integrating network and language data can predict complex, rare events like suicidality. We are confident we can apply this approach to predicting COVID-19 outbreaks in the near term. Future work would aim to generalize the model to predict dynamics of future contagions spreading via social networks. This model would give policymakers the capability to accurately target public health initiatives to vulnerable populations, conserving resources and limiting the spread of outbreaks at much earlier stages than current approaches allow. Interactive reports will allow users to compare different mode assumptions to forecast how specific interventions may shape the pandemic's trajectory.
