COVID-19 Case Forecasting
Using MATLAB, I tried to forecast the COVID-19 new forecast based on Compartmental Epidemiological Method (CEM) using US COVID case data from January 21 2020 to April 19 2021 sourced from Github
In CEM, we separate the population to 3 groups: Suspected, Infectious, and Recovered. Also, we assume that a constant rate of infection (β) and recovery (γ) per person as well as Recovered are no more Susceptible to the virus. After calculating the average constants then fitting the data, the predicted value do not represent the actual case well.
As the slope changes indefinitely over time, we instead used a rolling average of the short term β and γ value as well as exponential smoothing to integrate real data. This resulted in a better fit even though the error margin is still quite large.
