Glossary term
Glossary term
Governance and Compliance
A change in the relationship between inputs and outputs, even when input distributions look stable. Concept drift can quietly erode model relevance and decision quality. Concept drift is harder to detect than data drift and often requires outcome tracking, sampled human review, business-feedback loops, or external validation signals.
A shift in the relationship between features and the label. Over time, concept drift reduces a model's quality.
During training, the model learns the relationship between the features and their labels in the training set. If the labels in the training set are good proxies for the real-world, then the model should make good real world predictions. However, due to concept drift, the model's predictions tend to degrade over time.
For example, consider a binary classification model that predicts whether or not a certain car model is "fuel efficient." That is, the features could be:
car weight
engine compression
transmission type
while the label is either:
fuel efficient
not fuel efficient
However, the concept of "fuel efficient car" keeps changing. A car model labeled fuel efficient in 1994 would almost certainly be labeled not fuel efficient in 2024. A model suffering from concept drift tends to make less and less useful predictions over time.
Compare and contrast with nonstationarity.
Click the icon for additional notes.
Email spam classifiers face continuous concept drift as spammers change tactics, requiring frequent retraining at Google and Microsoft.
Credit risk concept drift was widely observed during 2008 to 2010 as default patterns shifted independently of borrower demographics.
Gama et al. (2014) A Survey on Concept Drift Adaptation remains a foundational reference for detection and adaptation strategies.
Created for this library
A bank's fraud team monitors concept drift weekly because fraudster tactics shift faster than the model retraining schedule.
A retail demand forecasting team triggers retraining whenever monitoring detects concept drift in shopper behavior after major holidays.
A subscription business watches concept drift on its churn model after pricing changes because the relationship between features and churn shifts.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License