Contributed talk

Speaker: Ye Chen (Northern Arizona University)

Location and Time: SAS 221A, Saturday, 11:15–11:45 AM

Title: Innovations in Flu Forecasting and Bayesian Algorithms

Abstract: The significance of infectious disease forecast in public health cannot be understated. Precise and timely forecasts can significantly enhance public health responses by guiding preparatory and mitigation strategies. Since 2013, CDC has been organizing the “FluSight Challenge” for every flu season to encourage academic and industry researchers to forecast national and regional flu activities in the United States. Currently, two primary real-time forecasting models dominate the field: statistical time series models and compartmental models. Statistical time series models are often sensitive to immediate changes, whereas compartmental models offer good long-term forecasting capabilities. Since the 2021 flu season, I have been leading a team named 'Los Alamos - NAU' in participating in the FluSight Challenge, and we have achieved some success. We have been developing advanced Bayesian algorithms to fit a complex compartmental model of flu transmission that incorporates real-time hospitalization data. Our approach utilizes Bayesian inference to ascertain the posterior probability density function, or the parameter posterior. In addition to parameter estimation, we have focused on quantifying uncertainty in predicted trajectories of daily case reports.