Cost-sensitive Boosting for Concept Drift

Proposes a cost-sensitive variant of AdaBoost for learning under concept drift, selectively reweighting outdated source-domain instances so that a boosted ensemble adapts to an evolving target distribution.