Wikitopia
Data drift
conceptML Concept
Try in Playground →
Overview
Use casemonitoring changes in input data distribution over time
Also see
Alternative to
concept drift
Based onstatistical hypothesis testing
Knowledge graph stats
Claims76
Avg confidence91%
Avg freshness100%
Last updatedUpdated 5 days ago
Trust distribution
100% unverified
Governance
Not assessed
Contribute governance data →

Data drift

concept

Changes in input data distribution between training and production environments affecting model performance.

Compare with...

is type of

ValueTrustConfidenceFreshnessSources
machine learning concept○UnverifiedHighFresh1
machine learning monitoring concept○UnverifiedHighFresh1

field of study

ValueTrustConfidenceFreshnessSources
machine learning○UnverifiedHighFresh1

category

ValueTrustConfidenceFreshnessSources
machine learning monitoring concept○UnverifiedHighFresh1

causes problem

ValueTrustConfidenceFreshnessSources
model performance degradation○UnverifiedHighFresh1

primary use case

ValueTrustConfidenceFreshnessSources
monitoring changes in input data distribution over time○UnverifiedHighFresh1
detecting changes in input data distribution over time in machine learning systems○UnverifiedHighFresh1
detecting changes in input data distribution that may degrade machine learning model performance○UnverifiedHighFresh1
monitoring changes in input data distribution compared to training data○UnverifiedHighFresh1
detecting changes in statistical properties of input data over time in machine learning systems○UnverifiedHighFresh1

impacts

ValueTrustConfidenceFreshnessSources
model performance○UnverifiedHighFresh1

causes problem for

ValueTrustConfidenceFreshnessSources
machine learning model performance○UnverifiedHighFresh1

integrates with

ValueTrustConfidenceFreshnessSources
model monitoring systems○UnverifiedHighFresh1
MLOps platforms○UnverifiedModerateFresh1
feature stores○UnverifiedModerateFresh1

occurs in domain

ValueTrustConfidenceFreshnessSources
machine learning operations○UnverifiedHighFresh1
production machine learning systems○UnverifiedHighFresh1

commonly affects

ValueTrustConfidenceFreshnessSources
production machine learning systems○UnverifiedHighFresh1

requires

ValueTrustConfidenceFreshnessSources
baseline data distribution○UnverifiedHighFresh1
reference dataset○UnverifiedHighFresh1
baseline reference data○UnverifiedHighFresh1
continuous data collection○UnverifiedHighFresh1
statistical monitoring techniques○UnverifiedHighFresh1

part of

ValueTrustConfidenceFreshnessSources
MLOps practices○UnverifiedHighFresh1
MLOps pipeline○UnverifiedHighFresh1

based on

ValueTrustConfidenceFreshnessSources
statistical hypothesis testing○UnverifiedHighFresh1

mitigation strategy

ValueTrustConfidenceFreshnessSources
model retraining○UnverifiedHighFresh1
online learning○UnverifiedHighFresh1

related concept

ValueTrustConfidenceFreshnessSources
concept drift○UnverifiedHighFresh1

supports model

ValueTrustConfidenceFreshnessSources
supervised learning models○UnverifiedHighFresh1
unsupervised learning models○UnverifiedHighFresh1

monitoring approach

ValueTrustConfidenceFreshnessSources
continuous data monitoring○UnverifiedHighFresh1

detects

ValueTrustConfidenceFreshnessSources
statistical changes in feature distributions○UnverifiedHighFresh1

related to

ValueTrustConfidenceFreshnessSources
concept drift○UnverifiedHighFresh1

detection method

ValueTrustConfidenceFreshnessSources
statistical hypothesis testing○UnverifiedHighFresh1
Kolmogorov-Smirnov test○UnverifiedModerateFresh1

monitored by

ValueTrustConfidenceFreshnessSources
AWS SageMaker Model Monitor○UnverifiedHighFresh1
TensorFlow Data Validation○UnverifiedModerateFresh1
Evidently AI○UnverifiedModerateFresh1
Whylabs○UnverifiedModerateFresh1

requires technique

ValueTrustConfidenceFreshnessSources
baseline data distribution establishment○UnverifiedHighFresh1

part of discipline

ValueTrustConfidenceFreshnessSources
MLOps○UnverifiedHighFresh1

measured using

ValueTrustConfidenceFreshnessSources
statistical distance metrics○UnverifiedHighFresh1
KL divergence○UnverifiedModerateFresh1
Kolmogorov-Smirnov test○UnverifiedModerateFresh1
population stability index○UnverifiedModerateFresh1
Wasserstein distance○UnverifiedModerateFresh1

detected by

ValueTrustConfidenceFreshnessSources
statistical tests○UnverifiedHighFresh1
Kolmogorov-Smirnov test○UnverifiedModerateFresh1
Population Stability Index○UnverifiedModerateFresh1

also known as

ValueTrustConfidenceFreshnessSources
dataset shift○UnverifiedHighFresh1
covariate shift○UnverifiedModerateFresh1

causes

ValueTrustConfidenceFreshnessSources
model performance degradation○UnverifiedHighFresh1

supports protocol

ValueTrustConfidenceFreshnessSources
Kolmogorov-Smirnov test○UnverifiedModerateFresh1
Population Stability Index○UnverifiedModerateFresh1

can trigger

ValueTrustConfidenceFreshnessSources
model retraining○UnverifiedModerateFresh1

triggers

ValueTrustConfidenceFreshnessSources
model retraining○UnverifiedModerateFresh1
model retraining workflows○UnverifiedModerateFresh1

monitored by platform

ValueTrustConfidenceFreshnessSources
Evidently AI○UnverifiedModerateFresh1

can be detected using

ValueTrustConfidenceFreshnessSources
statistical hypothesis testing○UnverifiedModerateFresh1

alternative to

ValueTrustConfidenceFreshnessSources
concept drift○UnverifiedModerateFresh1

can be measured with

ValueTrustConfidenceFreshnessSources
Kolmogorov-Smirnov test○UnverifiedModerateFresh1

Alternatives & Similar Tools

concept drift
alternative to
Compare →

Commonly Used With

model monitoring systemsMLOps platformsfeature stores

Related entities

Claim count: 76Last updated: 4/5/2026Edit history