SoNEA: Sensing Online Novelty Using Event Archives

Event detection and consequently novelty detection on time series data has recently attracted increasing attention from the computing community. In this paper, we describe a system that can detect novel events in wireless sensor networks termed SoNEA (sensing online novelty using event archives). This system is able to receive and process sensory data from sensor networks and dynamically detect novel events using intelligent novelty detection techniques. The detection is based on clustering techniques combined with cognitively motivated habituation theory. A novel scheme to predict missing values of sensor readings is also proposed based on this system. The results of the experiments exhibiting the performance of the proposed solution in detecting novel events and missing value predication are reported