Efficient quantile tracking using an oracle

Abstract Concept drift is a well-known issue that arises when working with data streams. In this paper, we present procedure allows quantile tracking to cope concept drift. We suggest using expected loss, popular loss function in regression, monitor the error, which, turn, used efficiently adapt The suggested procedures drift, and performance close theoretically optimal. were further applied three real-life streaming sets related Twitter event detection, activity recognition, stock trading. results show are efficient at adapting thereby documenting real-world applicability of procedures. asymptotic theory from statistics appealing theoretical property that, if stream distribution stationary over time, converge true quantile.