Real-time Anomaly Detection for Multivariate Data Streams
This work addresses the problem of detecting anomalies in streaming data for applications like monitoring systems, though it appears incremental as it builds on existing PEWMA methods.
The authors tackled real-time anomaly detection in multivariate data streams by developing an algorithm based on Probabilistic Exponentially Weighted Moving Average (PEWMA), which is resilient to various data shifts and works unsupervised, achieving good performance without requiring labeled examples.
We present a real-time multivariate anomaly detection algorithm for data streams based on the Probabilistic Exponentially Weighted Moving Average (PEWMA). Our formulation is resilient to (abrupt transient, abrupt distributional, and gradual distributional) shifts in the data. The novel anomaly detection routines utilize an incremental online algorithm to handle streams. Furthermore, our proposed anomaly detection algorithm works in an unsupervised manner eliminating the need for labeled examples. Our algorithm performs well and is resilient in the face of concept drifts.