The Forecasting and Case Study Modeling of COVID-19 in Chicago: A Data-driven Approach

Abstract

Probabilistic graphical models and machine learning are powerful data-driven tools for extracting useful knowledge from historical data; this knowledge can facilitate improved decision-making. With the prevailing efforts to combat the coronavirus disease 2019 (COVID-19) pandemic, there are still uncertainties that are yet to be discovered about its spread, future impact, and resurgence. In this paper, a data-driven approach has been adopted in distilling the hidden information about COVID-19 and its symptoms. This paper proposes: a Bayesian network which encodes the causal relationships among COVID-19 symptoms, an unsupervised machine learning algorithm that learns symptoms pattern in COVID-19 dataset, a deep neural network which predicts the symptoms class of patients based on clustering experience with a 99.47% testing accuracy, and a time-series forecasting model that predicts the trend of COVID-19. The results from the experiments show the capability of data-driven methods in addressing the concerns of the society and government in understanding the uncertainties about the virus, providing insights on developing policies, and reducing the spread of the virus.