Module drift 

Data drift detection for ML pipelines

Detects distribution changes between dataset versions or time periods. Implements Jidoka—building quality in at the data layer before training.

Example

use alimentar::drift::{DriftDetector, DriftTest};

let detector = DriftDetector::new(reference_dataset)
    .with_test(DriftTest::KolmogorovSmirnov)
    .with_test(DriftTest::PSI)
    .with_alpha(0.05);

let report = detector.detect(&current_dataset)?;
if report.drift_detected {
    println!("Drift detected in columns: {:?}", report.drifted_columns());
}

Structs

ColumnDrift

Per-column drift result

DriftDetector

Statistical drift detector

DriftReport

Overall drift detection report

Enums

DriftSeverity

Severity of detected drift

DriftTest

Statistical tests for drift detection