Concept drift is the phenomenon whereby the statistical properties of a target variable change over time relative to the variables used to predict it, degrading model performance. In practice it manifests as covariate shift (the input distribution P(X) changes while the conditional P(y|X) stays fixed), prior shift (P(y) changes but P(X|y) stays fixed), or true concept drift (the mapping P(y|X) itself changes). Detection methods compare distributions of model inputs or residuals across time windows using statistical tests (KS, MMD, ADWIN). The persistence mechanism is institutional: monitoring pipelines and retraining protocols maintain the concept in production systems. [formal: conceptus-errantis | substrate: mind | horizon: hours | explicit: yes | epoch: 0.01]
Accepted ontology entry
concept drift
Concept drift is the phenomenon whereby the statistical properties of a target variable change over time relative to the variables used to predict it, degrading model performance. In practice it manifests as covariate shift (the input dist…
Definition
Why it is in scope
The phenomenon in machine learning and statistics where the statistical properties of the target variable or input data change over time, causing a model trained on older data to degrade in predictive performance. It is human-made, designed as a diagnostic concept in model operations, and persists through the practice of data science, MLOps tooling, and sustained research into non-stationary learning environments.
Names and aliases
- concept driften · CANONICAL
Relations from this entry
- cmrxj3acr03cmsoacx73fal1oDERIVED_FROM →
Concept drift originated in statistical process control (1920s Shewhart charts) and was later adopted by machine learning. Statistics predates ML as a discipline, making it the historical source.
- cmsdvrauc04by3vv36su3q9jqDERIVED_FROM →
Time series analysis existed first and fed into the concept of concept drift — the idea that data distributions change over time builds on the temporal ordering inherent in time series. Which came first? Time series as a field predates concept drift; concept drift extends time series analysis to address distribution shifts within temporal data.
Relations to this entry
No accepted relations in this direction.
Record identity
- Created
- Aug 3, 2026, 8:46 AM UTC
- Content hash
- 6efd4658516d4c122849ebd64c72c08f630a49b332c0c262e160735de9ff0da4