Merging Two Arima Models for Energy Optimization in WSN

TL;DR

This paper proposes a method to optimize energy consumption in wireless sensor networks by merging ARIMA models for predictive data collection, reducing unnecessary data transmissions while maintaining accuracy.

Contribution

It introduces a two-phase approach that combines ARIMA modeling with adaptive data transmission to improve energy efficiency in sensor networks.

Findings

Significant reduction in data transmissions achieved.

Maintained prediction accuracy within predefined error bounds.

Enhanced energy efficiency in sensor data collection.

Abstract

Use of ARIMA model in Sensor network The basic idea of our energy efficient information collection scheme is to suppress data transmission if the data sampled by sensor nodes are predictable by the sink node. This is done in two phases 1) Preliminary Data Collection- During this phase sink node collects enough data so that it can build up ARIMA model for each node. Then sink node selects a model for the particular node and sends back the corresponding model parameters to the node and also keeps them with it. Selecting the model for a node there is a tradeoff between energy consumption and accuracy of prediction. So we choose the model according to C = {\alpha} xMAE + (1 - {\alpha}) x rtran 0=< {\alpha} =<1 where the model should minimize C. Here MAE is Mean Absolute Error which is normalized by some predefined error tolerance and rtran is the ratio of number of samples transmitted over total number of samples. 2) Adaptive Data Collection- After the sensor node has received the model parameters it checks each actual data value with the data value calculated from the parameters received. If there is deviation beyond some predefined error tolerance then only it sends the original data value to the sink node.