Real-Time Anomaly Detection for Multivariate Data Stream

Anonymous

Keywords: machine-learning, signal-processing

Abstract: The paper titled “Probabilistic reasoning for streaming anomaly detection” from MIT CSAIL proposed a framework for performing online anomaly detection on univariate data. Unfortunately, most of the data in the real world are multivariate. Hence, mandating the need for more research into performing online anomaly detection in multivariate data. We have been inspired by their work and extended their framework to support multivariate data with some clever optimizations to build a scalable system.