Repository navigation
Integrate multiseries into AutoMLSearch - #4270
Conversation
Codecov Report
@@ Coverage Diff @@
## main #4270 +/- ##
=======================================
+ Coverage 99.7% 99.7% +0.1%
=======================================
Files 355 355
Lines 38959 39073 +114
=======================================
+ Hits 38838 38953 +115
+ Misses 121 120 -1
|
| random_seed: Union[int, float] = 0, | ||
| maxiter: int = 10, | ||
| use_covariates: bool = True, | ||
| use_covariates: bool = False, |
There was a problem hiding this comment.
Flipped this default for the sake of speed. My test example did not train within a 5 minute window when use_covariates was True, but it ran in <10 seconds when use_covariates was False.
There was a problem hiding this comment.
I think it would be nice to see performance tests for use_covariates turned on or off. Could possibly turn it off only for tests if performance is greatly improved with covariates!
jeremyliweishih
left a comment
There was a problem hiding this comment.
LGTM and agreed with @chukarsten on potentially refactoring is_multiseries into a separate problem type!
| def get_estimators(problem_type, model_families=None, excluded_model_families=None): | ||
| def _filter_multiseries_estimators(estimators, is_multiseries): | ||
| if is_multiseries: | ||
| return [estimator for estimator in estimators if estimator.is_multiseries] |
There was a problem hiding this comment.
nit and maybe for a follow up: could estimator.is_multiseries be something like estimator.supports_multiseries? Think its more clear especially since we're passing is_multiseries everywhere for now.
| random_seed: Union[int, float] = 0, | ||
| maxiter: int = 10, | ||
| use_covariates: bool = True, | ||
| use_covariates: bool = False, |
There was a problem hiding this comment.
I think it would be nice to see performance tests for use_covariates turned on or off. Could possibly turn it off only for tests if performance is greatly improved with covariates!
* Add multiseries time series regression as problem type * Completely revamp to multiseries based on problem type
| f"Dataset size is too small to create holdout set. Minimum dataset size is {self._HOLDOUT_SET_MIN_ROWS} rows, X_train has {len(X_train)} rows. Holdout set evaluation is disabled.", | ||
| ) | ||
|
|
||
| # For multiseries problems, we need to mke sure that the data is primarily ordered by the time_index rather than the series_id |
There was a problem hiding this comment.
Nit: we need to mke sure -> we need to make sure
Closes #4266