How to choose a multi-step forecasting strategy?
Let’s summarize all the different strategies that we learned now in a table:
Figure 17.9 – Multi-step forecasting strategies – a summary
Here, the following applies:
- S.O: Single output
- M.O: Multi-output
and 
: Training and inferencing time of a single output model
and 
: Training and inferencing time of a multi-output model (practically, 
is larger than 
mostly because multi-output models are typically DL models and their training time is higher than standard ML models)
: The horizon
, where 
is the number of blocks in the IBD strategy
is some positive real number
The table will help us understand and decide which strategy is better from multiple perspectives:
- Engineering complexity: Recursive, Joint, RecJoint << IBD << Direct, DirRec << Rectify
- Training time: Recursive << Joint (typically
) << RecJoint <...
