Evaluation of Deep Learning Methods in a Dual Prediction Scheme to Reduce Transmission Data in a WSN

Carlos R Morales; Fernando Rangel de Sousa; Valner Brusamarello; Nestor C Fernandes

doi:10.3390/s21217375

Evaluation of Deep Learning Methods in a Dual Prediction Scheme to Reduce Transmission Data in a WSN

Sensors (Basel). 2021 Nov 6;21(21):7375. doi: 10.3390/s21217375.

Authors

Carlos R Morales¹, Fernando Rangel de Sousa¹, Valner Brusamarello², Nestor C Fernandes³

Affiliations

¹ Department of Electrical and Electronic Engineering, Universidade Federal de Santa Catarina, Florianópolis 88040-900, Brazil.
² Department of Electrical Engineering, Universidade Federal do Rio Grande do Sul, Porto Alegre 91501-970, Brazil.
³ Traceback Technologies, Rua Antônia dos Santos Silveira, Florianópolis 88090-145, Brazil.

Abstract

One of the most important challenges in Wireless Sensor Networks (WSN) is the extension of the sensors lifetime, which are battery-powered devices, through a reduction in energy consumption. Using data prediction to decrease the amount of transmitted data is one of the approaches to solve this problem. This paper provides a comparison of deep learning methods in a dual prediction scheme to reduce transmission. The structures of the models are presented along with their parameters. A comparison of the models is provided using different performance metrics, together with the percent of points transmitted per threshold, and the errors between the final data received by Base Station (BS) and the measured values. The results show that the model with better performance in the dataset was the model with Attention, saving a considerable amount of data in transmission and still maintaining a good representation of the measured data.

Keywords: energy saving; forecasting; neural networks; wireless sensor networks.