Boosting classifiers for drifting concepts

In many real-world classification tasks, data arrives over time and the target concept to be learned from the data stream may change over time. Boosting methods are well-suited for learning from data streams, but do not address this concept drift pro