Design - comparison MapieTimeSeriesRegressor
and sktime
APIs
#379
Labels
Enhancement
Type: enhancement (new feature or request)
Needs decision
The MAPIE team is deciding what to do next.
Other or internal
If no other grey tag is relevant or if issue from the MAPIE team
Source: developers
Proposed by developers.
As previously discussed, a quick review of
MapieTimeSeriesRegressor
againstsktime
APIs.capability:pred_int
(can predict probabilistically), andignores-exogeneous-X: False
, as it uses exogenous datafit
andpredict
map onto each other, with slightly different args.mapie
'spartial_fit
issktime
'supdate
method.predict
(ifalpha
is passed) issktime
'spredict_interval
.X
is absolutely required inmapie
(is that true?), whereas theoretically the implemented estimator does not need it, andsktime
also assumes that every forecaster can be run withoutX
, withy
onlysktime
sktime
,fit
andpredict
cannot be given additional parameters, all non-data/task arguments should be in the constructor. So,ensemble
oroptimize_beta
should move to the constructor, frompredict
.From an interfacing perspective, using the maximal extension template with filling only
predict_interval
should work: https://github.com/sktime/sktime/blob/main/extension_templates/forecasting.py(possibly also filling in
predict_proba
with anEmpirical
distribution)The text was updated successfully, but these errors were encountered: