-
Notifications
You must be signed in to change notification settings - Fork 861
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
[timeseries] Implement recurrent forecasting model based on TabularPredictor and MLForecast #3177
Conversation
Job PR-3177-eeac00a is done. |
Job PR-3177-a1ee41c is done. |
@@ -38,6 +39,7 @@ | |||
AutoETS=AutoETSModel, | |||
AutoARIMA=AutoARIMAModel, | |||
DynamicOptimizedTheta=DynamicOptimizedThetaModel, | |||
RecursiveTabular=RecursiveTabularModel, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
which presets are we intended to add for this model?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I plan to update these in a follow-up PR.
) -> TimeSeriesDataFrame: | ||
predictions = self._predict_without_quantiles(data, known_covariates) | ||
for q in self.quantile_levels: | ||
predictions[str(q)] = predictions["mean"] + self.quantile_adjustments[q] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Wondering how the quantile adjustment is done
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Here we just use the empirical distribution of the quantiles to generate dummy quantiles. I will add another tree-based model that is trained to predict the quantiles using the wQL loss in a future PR.
Description of changes:
RecursiveTabularModel
that uses theTabularPredictor
to forecast time series 1 step into the future. The overall forecast is obtained by unrolling the model predictions over time.By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.