🌊 Covariate Drift (a.k.a. Covariate Shift)

Covariate Drift (a.k.a. Covariate Shift)#

A change in the input distribution while the input-output relationship holds.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

Covariate drift (covariate shift) is a change in the distribution of the input features a model sees — the production inputs no longer look like the training inputs — while the feature-to-label rule stays the same:

\[p_{\text{train}}(x) \neq p_{\text{prod}}(x), \qquad p(y \mid x)\ \text{unchanged}.\]

The model is being asked about a different population than it learned on.

How it differs#

It is one of three dataset shifts. Covariate drift moves p(x) (the inputs), label drift moves p(y) (the target mix), and concept drift moves p(y | x) (the relationship itself). Only concept drift changes the rule; covariate drift changes who you’re scoring.

Detecting and fixing it#

It is caught by comparing feature distributions per column with PSI or the KS test. Remedies include importance weighting — reweighting training points by the density ratio \(w(x) = p_{\text{prod}}(x) / p_{\text{train}}(x)\) — retraining on recent data, and building robust features (winsorized, log-scaled, sensible bins).

Theme: Distribution Shift & Drift · All terminology

Hint

Mind map — connected ideas

Label Drift (a.k.a. Target Drift) · Concept Drift · Dataset Shift · Data Drift · Drift Detection · PSI (Population Stability Index)

Hint

More in Distribution Shift & Drift

Cardinality in Categorical Data · Categorical Drift · Categorical Explosions · Classifier Two-Sample Tests (C2STs) · Concept Drift · Data Drift · Dataset Shift · Drift Detection · Drift Guardrails · Energy Distance · Jensen–Shannon (JS) Divergence · KS shift (Kolmogorov–Smirnov shift) · Kullback–Leibler (KL) Divergence · Label Drift (a.k.a. Target Drift)