-
Use the SEIR model from CoronaTracker: World-wide COVID-19 Outbreak Data Analysis and Prediction
-
Fix two model parameters from that paper as:
- incubation rate, sigma = 1 / (5.2 days)
- recovery rate, gamma = 1 / (2.3 days)
-
Estimate model variables (
S,E,I,R
) and model parameterbeta
, which controls the rate of spread. -
The observed data is a state's number of infected patients each day.
-
Use an unscented Kalman filter with default parameter settings.
-
After filtering, predict for 30 days into the future.
-
Identify the date with peak infections.
- Define
N
as state's population (from Wikipedia) - The model breaks down at long time predictions (i.e., populations < 0).