Hierarchical Regression for Data Acquisition in Wireless Sensor Networks

In this paper, the regression framework is proposed to hierarchically organize the three regression procedures, that is, local data modelling, cluster data modelling and global data modelling, for the energy-efficient data acquisition in wireless sensor networks (WSNs). There are two advantages arising from our method. First, there is no need for any transmissions of the measured data in the modelling procedures, which leads to significant energy savings, even for inhomogeneous networks. Second, there is no error peaking phenomenon occurred at cluster boundaries, which contributes to considerable performance improvement.