Adaptive Event Forecasting in Wireless Sensor Networks

Since the emergence of the earliest wireless sensor networks (WSNs), a great deal of effort has been made to support target tracking applications. Although target tracking, or other services such as dynamic sleep scheduling or early warning systems require some level of event forecasting, there was little research on this topic, in spite of its importance. In this paper we formalize a framework for event forecasting based on fuzzy logic and propose an adaptive and scalable algorithm to predict events in the network. Moreover, for each forecast our solution provides a bounded confidence level based on learned time-space fuzzy signatures (TSS). We evaluate the proposed method by means of simulations and provide parameter sensitivity analysis as well.