Since a comparsion testbed was created in #8 , that can be used to identify low-hanging-fruit improvements.
See if higher forecasting accuracy can be achieved by tuning:
hyperparameters
loss functions
model types
Acceptence criteria:
an analyses is documented on how the forecasts can be improved by minor model/pipeline adjustments
Repeat for both cases
Should this be further explored in the Full Project?
Since a comparsion testbed was created in #8 , that can be used to identify low-hanging-fruit improvements. See if higher forecasting accuracy can be achieved by tuning:
Acceptence criteria: