Nonparametric and Online Change Detection in Multivariate Datastreams Using QuantTree

We address the problem of online change detection in multivariate datastreams, and we introduce QuantTree Exponentially Weighted Moving Average (QT-EWMA), a nonparametric change-detection algorithm that can control expected time before false alarm, yielding desired Run Length (ARL$_0$). Controlling alarms is crucial many applications rarely guaranteed by algorithms monitor datastreams without knowing data distribution. Like algorithms, QT-EWMA builds model distribution, our case histogram, from stationary training set. To even when set extremely small, propose QT-EWMA-update, which incrementally updates histogram during monitoring, always keeping ARL$_0$ under control. Our experiments, performed on synthetic real-world demonstrate QT-EWMA-update alarm rate better than state-of-the-art methods operating similar conditions, achieving lower or comparable delays.