An Adaptive Prediction Strategy with Clustering in Wireless Sensor Network

In Wireless Sensor Network, sensed data reflects two types of correlations of physical attributes: spatial and temporal. In this paper, a scheme named, Adaptive Prediction Strategy with ClusTering (APSCT) is proposed. In APSCT, a data-driven clustering and grey prediction model is used to exploit both the correlations. APSCT minimizes the transmission of messages in the network. However, the use of prediction includes additional computation overhead. There is a trade-off between prediction accuracy and energy consumption in computation and communication in wireless networks. This paper also gives an approach to calculate the upper and lower bound of the prediction interval which is used to evaluate different confidence levels and provides an energy-efficient sensor environment. Simulation is carried out on real-world data collected by Intel Berkeley Lab and results are compared with existing approaches.