Open ParadaCarleton opened 11 months ago
In principle this sounds like a good idea. However, we do have UnivariateTimeTypeToContinuous
and I would prefer to avoid code duplication.
Perhaps we could add date_time_encoder=UnivariateTimeTypeToContinous()
as a new ContinuousEncoder
hyperparameter, and use that to treat columns with scitype ScientificTimeType
(corresponding to any machine type <: Dates.TimeType).
What do you think?
Sounds good!
Right now, inputting a date into a
ContinuousEncoder
just drops it, but it seems more reasonable to handle by converting dates into a number of days.