SARIMAX Modeling

Pilot modeling using Los Angeles

model validation using test data

the model is not able to predict trend changes, so the difference between our forecast and the actual data widens as time goes on

however, unless we incorporate multiple factors that are used by actual forecasting models, such as:

change in R-values, % infected, % recovered with immunity, county-by-county lockdown recommendations, and other demographics (age, race, ethnicity, income, employment)

forecasting the direction change is very difficult

model using all data points

Modeling for 7 Counties Highest New Cases/100k

Tehama County

Riverside County

Colusa County