Teach AutoARIMAModel to work with out-sample predictions
#830
Conversation
…m_steps, fix tests
|
🚀 Deployed on https://deploy-preview-830--etna-docs.netlify.app |
Codecov Report
@@ Coverage Diff @@
## master #830 +/- ##
==========================================
+ Coverage 84.64% 84.65% +0.01%
==========================================
Files 130 130
Lines 7455 7409 -46
==========================================
- Hits 6310 6272 -38
+ Misses 1145 1137 -8
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
| @@ -45,126 +41,11 @@ def __init__( | |||
| Training parameters for auto_arima from pmdarima package. | |||
| """ | |||
| self.kwargs = kwargs | |||
| self._model: Optional[ARIMA] = None | |||
| self.regressor_columns: List[str] = [] | |||
| super().__init__() | |||
There was a problem hiding this comment.
This should be before self.kwargs I guess.
There was a problem hiding this comment.
Why? In class AutoARIMAModel(PerSegmentPredictionIntervalModel) we set self.kwargs before calling super constructor.
There was a problem hiding this comment.
By mistake, no?
It seems logic in parent class could reset self and so it's better to make __init__ call before
There was a problem hiding this comment.
And vise-versa, logic in current class can reset the logic in the base class and broke smth.
etna/models/sarimax.py
Outdated
| "Try to encode this columns manually." | ||
| ) | ||
|
|
||
| def get_model(self) -> SARIMAX: |
There was a problem hiding this comment.
It seems we return not Sarimax
Before submitting (must do checklist)
Proposed Changes
Look #790.
Closing issues
Closes #790.