Load Reduction Using Temporal Modeling and Prediction in Periodic Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) operate in an energy constrained environment, and judicious use of limited battery of sensor nodes is a priority. Load reduction aids in prudent usage of battery by reducing the amount of data transmitted across the network without loss in underlying information, thereby increasing the network lifetime. This paper showcases a load reduction technique where we understand patterns in temporal data and create an adaptive prediction model based on M5P algorithm in the WEKA toolkit. The model predicts measurements, and only when sensor measurements do not agree the predictions, sensor nodes send data to the sink. This brings down the amount of data transmitted, leading to reduced communication and energy consumption. Preliminary results indicate 70% reduction in data transmission across the network, proving the efficacy of the temporal modeling in reducing amount of data sent, consequently saving energy, and improving the network lifetime.