dataeval.shift.DriftWasserstein.Stats¶

class dataeval.shift.DriftWasserstein.Stats¶

Per-feature statistics from Wasserstein drift detection.

ratio¶

Mean distance ratio across all features. Values substantially above 1.0 indicate drift; values near 1.0 indicate no drift.

Type:¶

float

feature_drift¶

Boolean array indicating which features show drift. Shape matches the number of features in the input data.

Type:¶

NDArray[bool]

feature_ratios¶

Distance ratio for each feature (operational / baseline). Shape matches the number of features in the input data.

Type:¶

NDArray[np.float32]

distances¶

Wasserstein distances between training and operational data, one per feature. Shape matches the number of features.

Type:¶

NDArray[np.float32]

baseline_distances¶

Wasserstein distances between training and validation data, one per feature. Computed once during fit() and reused across all subsequent predict() calls. Shape matches the number of features.

Type:¶

NDArray[np.float32]