Anomaly Detection via Real-Time Monitoring of High-Dimensional Event Data

Modern technological developments, such as smart chips, sensors, and wireless networks, have revolutionized data-collection processes. One type of data that can be highly beneficial is event data due to the general conceptualization of an event. Monitoring event data enables real-time monitoring since an observation becomes available as soon as the event happens. Most of the available literature on monitoring event data is focused on vector-based time between events (TBEs) data. Methods for this type of data incorporate monitoring delays either due to overseeing temporal dependencies between variables or the need to wait until a complete vector is observed. To tackle these issues, we propose a multivariate monitoring scheme of event data that signals in real time. Our contribution is twofold: Our proposed method can monitor high-dimensional event data and it is computationally quick and easy to implement, thereby, outperforming existing methods that are only feasible to implement up to ten dimensions.