[timeseries] add feature importance to TimeSeriesPredictor #4033
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
canerturkmen
added
enhancement
Contributor
|
shchur
reviewed
Apr 3, 2024
Collaborator
shchur
left a comment
shchur
left a comment
There was a problem hiding this comment.
This looks very elegant!
The main points that I want to discuss:
- Moving all feature importance logic to Trainer
- Using statistics of train data for masking
- Use full train_data/tuning_data by default
- Fixing seed for determinism
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
shchur
reviewed
Apr 3, 2024
| ) | ||
|
|
||
| if features is None: | ||
| features = feature_importance_transform.covariate_metadata.all_features |
Collaborator
There was a problem hiding this comment.
It's possible that some features are present in data.columns or data.static_features.columns but are missing from covariate_metadata.all_features because TimeSeriesFeatureGenerator removes non-informative features from the data (e.g., duplicate feature, feature consisting of all constant values, feature that takes different value for each row, unrecognized dtype time datetime).
We should manually check the columns in original data and assign feature importance of 0 to columns that are missing from transformed data.
Contributor
There was a problem hiding this comment.
This is what is done in Tabular I believe
Contributor
|
Innixma
reviewed
Apr 3, 2024
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
Contributor
Author
Thanks for the review @Innixma ! This was the first cut trying to pin down the design. I'll base my replies on that. |
Contributor
|
shchur
reviewed
Apr 4, 2024
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
Contributor
|
Contributor
|
Contributor
|
1 similar comment
Contributor
|
Contributor
|
shchur
reviewed
Apr 5, 2024
timeseries/src/autogluon/timeseries/models/autogluon_tabular/mlforecast.py
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py
Outdated
Show resolved
Hide resolved
Contributor
|
canerturkmen
force-pushed
the
ts-feature-importance-transforms
branch
from
c517448 to
68b8095
Compare
Contributor
|
shchur
approved these changes
Apr 5, 2024
Collaborator
shchur
left a comment
shchur
left a comment
There was a problem hiding this comment.
LGTM, thank you for taking care of this massive feature! 🚀
timeseries/src/autogluon/timeseries/models/abstract/abstract_timeseries_model.py
Show resolved
Hide resolved
Contributor
|
LennartPurucker
pushed a commit
to LennartPurucker/autogluon
that referenced
this pull request
Jun 1, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issue #, if available:
#3924
Description of changes:
This draft PR introduces the high level design for
TimeSeriesPredictor.feature_importance.Additional to-dos:
(stretch) Revisit prediction caching logic to prevent models that don't use features from re-inferringBy submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.