Add MissingCounter metric
#520
Conversation
| @@ -296,4 +296,55 @@ def wape(y_true: ArrayLike, y_pred: ArrayLike, multioutput: str = "joint") -> Ar | |||
| return np.sum(np.abs(y_true_array - y_pred_array), axis=axis) / np.sum(np.abs(y_true_array), axis=axis) # type: ignore | |||
|
|
|||
|
|
|||
| __all__ = ["mae", "mse", "msle", "medae", "r2_score", "mape", "smape", "sign", "max_deviation", "rmse", "wape"] | |||
| def count_missing_values(y_true: ArrayLike, y_pred: ArrayLike, multioutput: str = "joint") -> ArrayLike: | |||
There was a problem hiding this comment.
I haven't added it into etna/metrics/__init__.py. Do we have to do it?
There was a problem hiding this comment.
I dont think it is needed there.
|
🚀 Deployed on https://deploy-preview-520--etna-docs.netlify.app |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #520 +/- ##
=======================================
Coverage 90.32% 90.32%
=======================================
Files 256 256
Lines 17199 17214 +15
=======================================
+ Hits 15535 15549 +14
- Misses 1664 1665 +1 ☔ View full report in Codecov by Sentry. |
| @@ -296,4 +296,55 @@ def wape(y_true: ArrayLike, y_pred: ArrayLike, multioutput: str = "joint") -> Ar | |||
| return np.sum(np.abs(y_true_array - y_pred_array), axis=axis) / np.sum(np.abs(y_true_array), axis=axis) # type: ignore | |||
|
|
|||
|
|
|||
| __all__ = ["mae", "mse", "msle", "medae", "r2_score", "mape", "smape", "sign", "max_deviation", "rmse", "wape"] | |||
| def count_missing_values(y_true: ArrayLike, y_pred: ArrayLike, multioutput: str = "joint") -> ArrayLike: | |||
There was a problem hiding this comment.
I dont think it is needed there.
| @property | ||
| def greater_is_better(self) -> None: | ||
| """Whether higher metric value is better.""" | ||
| return None |
There was a problem hiding this comment.
It is understandable that we can't control missing values in the true target, but shouldn't it be False? So we prefer more data over the less data.
There was a problem hiding this comment.
I think metrics should be used to compare different methods with each other on the same dataset. This metric will be the same for the same dataset because it doesn't depend on the model.
So, I'm not really sure what value should it have.
| "metric", | ||
| (MSE(mode="per-segment", missing_mode="ignore"),), | ||
| "metric, expected_type", | ||
| ((MSE(mode="per-segment", missing_mode="ignore"), type(None)), (MissingCounter(mode="per-segment"), float)), |
There was a problem hiding this comment.
There is types.NoneType, but it was added in python 3.10: https://docs.python.org/3/library/types.html#types.NoneType
# Conflicts: # CHANGELOG.md
Before submitting (must do checklist)
Proposed Changes
See #517.
Closing issues
Closes #517.