Driftsteams
Driftsteams is a term used in data science to describe groups of drift phenomena that occur in non-stationary data streams. A driftsteam is conceived as a coherent cluster of distribution changes that appear within a bounded time window and share common drivers, such as a change in underlying processes, sensor bias, or external shocks. The concept emphasizes the connected nature of related drift events rather than treating each instance in isolation.
Origins and usage: The term driftsteam emerged in exploratory studies of concept drift to capture the idea
Definitions and components: A driftsteam comprises members (drift indicators) that occur close in time and share
Detection and analysis: Methods for detecting driftsteams combine concept-drift detectors with streaming clustering and change-point techniques.
Applications and limitations: Driftsteams are used in machine learning for monitoring model drift, in sensor networks