/seir-kalman-filter

Primary LanguageMATLAB

Estimate parameters of SEIR model using an unscented Kalman filter and state data.

  1. Use the SEIR model from CoronaTracker: World-wide COVID-19 Outbreak Data Analysis and Prediction

  2. Fix two model parameters from that paper as:

  • incubation rate, sigma = 1 / (5.2 days)
  • recovery rate, gamma = 1 / (2.3 days)

  1. Estimate model variables (S,E,I,R) and model parameter beta, which controls the rate of spread.

  2. The observed data is a state's number of infected patients each day. See Jupyter notebook.

  3. Use an unscented Kalman filter with default parameter settings.

  4. After filtering, predict for 30 days into the future.

  5. Identify the date with peak infections.

Notes

  • Define N as state's population (from Wikipedia)
  • The model breaks down at long time predictions (i.e., populations < 0).

Example output

alt text