Extracting Event Patterns from Telemetry Data

Abstract

Typical low-level telemetry displays provide little support for understanding how a monitored system is changing over time. Similarly, advanced display techniques have not explicitly addressed the representation of change and events. This paper describes a study which seeks to work towards a foundation for displays which highlight coherent units of behavior in system properties (“events”). In the study, 17 medical professionals viewed simulated telemetry from a surgical procedure. A technique adapted from social perception research was used to trace participants' flow of attention to new information as they viewed the case. The results demonstrate that observers dynamically construct higher-level descriptions of system behavior from instantaneous data values, and that these descriptions implicate various types of contextual information.