Guang Lin, Purdue University, and Ehsan Kharazmi, Brown University. Title: "Predicting the COVID-19 pandemic with uncertainties using data-driven models"
Guang Lin, Purdue University
Ehsan Kharazmi, Brown University
January 14, 2021, IMAG/MSM WG on Multiscale Modeling for Viral Pandemics
Grace Peng Slides
Grace Peng Video
Guang Lin and Ehsan Kharazmi Slides
Guang Lin and Ehsan Kharazmi Video
Guang Lin and Ehsan Kharazmi Abstract: We have developed an integer-order COVID-19 epidemic model and a fractional-order COVID-19 epidemic model to reconstruct and forecast the transmission dynamics of COVID-19 in New York City. To quantify the uncertainties in the proposed data-driven epidemic model, we have investigated model sensitivity analysis, structural and practical identifiability analysis, model calibration, and uncertainty quantification. We have employed Bayesian model calibration and physics-informed machine learning algorithms to calibrate the model parameters. In the early stage of the outbreak in New York City, the reproduction number was around 4.3, which indicates this outbreak has high transmissibility. We observed that multi-pronged interventions, such as the stay-at-home order and social distancing, had positive effects on controlling the outbreak and slowing the virus's spread. In addition, we employed the proposed data-driven models to evaluate the effects of various strategies to deploy the COVID-19 vaccine to control the pandemic. We have also applied the formulation to infer the dynamics of COVID-19 in other cities/states, where the spread dynamic is different from New York City.