New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[ENH] Tracking issue for make_reduction
outstanding issues
#5736
Comments
FYI @benHeid |
We have been trying to start using reduction forecasters for our use-case (including exogenous features I wonder why |
@mlisovyi, thanks for the feedback! Yes, the results might be interesting. In a nutshell, the biggest problem is lack of contributor time on these estimators, as fixing some of the known issues is unpleasant, and volunteers tend to gravitate elsewhere. The current situation:
Mostly a time investment issue. Would you perhaps be willing to help, if your use cases require it?
Yes, or womanpower, anyone who can help with the coding. |
This issue is to track work items related to the reducers in
forecasting.compose._reduce
.current reducer
simplification
skpro
) inmake_reduction
, direct reduction #5536_BaseWindowForecaster
does not seem consistent - this should be revisited.bugs
global
setting was not covered by tests, and fails tests, see here: [BUG] fix global pooling in forecasting reducers #5722X
is also lagged, that is unexpected for the user and should be changed. Best way is probablylag_X
parameter, then change the default via deprecation.improvements
sklearn
regressors support diagnostics, e.g., feat importances based on name. These are not being passed on.X
, behaviour in global case, cutting of windows, which parts of the time series are exactly used in fit/predict, etc)extensions
X
gets lagged or not, similar to the new direct reducerRelated issues or PR:
skpro
) inmake_reduction
, direct reduction #5536The text was updated successfully, but these errors were encountered: