dataeval.shift.DriftOutput¶

class dataeval.shift.DriftOutput¶

Output class for drift detector classes.

Provides common fields returned by all drift detection methods, containing instance-level drift predictions and summary statistics.

For non-chunked mode, details holds a detector-specific TypedDict with test statistics (including p_val). For chunked mode, details holds a polars.DataFrame with per-chunk results.

drifted¶

Whether drift was detected in the analyzed data. True indicates significant drift from reference distribution.

Type:¶

bool

threshold¶

Significance threshold used for drift detection, typically between 0 and 1. For multivariate methods, this is the corrected threshold after Bonferroni or FDR correction.

Type:¶

float

distance¶

Instance-level test statistic or distance metric. Typically >= 0, but can be slightly negative for metrics like unbiased MMD². For univariate methods, this is the mean distance across all features. Higher values indicate greater deviation from reference distribution.

Type:¶

float

metric_name¶

Name of the metric used (e.g., “mmd2”, “auroc”, “ks_distance”).

Type:¶

str

details¶

Detector-specific statistics (TypedDict) for non-chunked mode, or a polars.DataFrame of per-chunk results for chunked mode.

Type:¶

TDetails

data()¶

Return the output data as a dictionary.

Return type:¶

dict[str, Any]

meta()¶

Metadata about the execution of the function or method for the Output class.

Return type:¶

ExecutionMetadata