Open
Description
Feedback from Tianyue:
All the examples I see... don’t have very good prediction performance, although we discuss some reasons after the examples (like we don’t exclude the outliers). Maybe we can use one real-world prediction task to provide a thorough guide for people who are not experts in this area.
In other words, instead of giving people “bad” results and pointing out reasons afterwards, why don’t we just give people the “correct” procedure to follow?
We should not show the bad case forecasts. Maybe show some decent hospital admissions forecasts which are more relevant and more possible.