Multi-temporal-scale event detection and clustering in IoT systems

Abstract

Sensor-based IoT systems detect events from data and take appropriate actions through the event processing process. The core of event processing, event rules, is typically manually defined by domain experts. However, there are limitations to domain experts manually setting rules for all the unlabeled events in runtime of IoT systems. Therefore, there is a need for methods that support the generation of rules for unlabeled events. In this study, we address this issue by adding two phases to the existing event processing process. The first phase is the detection of unlabeled events from the raw data. Considering the characteristics of IoT systems, we propose MulTemS, an extension of anomaly detection techniques that can detect events of various temporal-scales. The second phase is the formation of clusters among similar events. We propose FeatCNC, which predicts the number of clusters through feature extraction and performs domain-independent clustering. Through experiments, we demonstrate that MulTemS can effectively detect events of multiple temporal-scales, and FeatCNC can reliably cluster events across diverse domains. Additionally, we verify that the integration of these two phases results in the better formation of clusters that capture the characteristics of the events.