Leveraging Fog Layer Data Prediction Using Deep Learning for Enhanced IoT Sensor Longevity
Made Adi Paramartha Putra, Mideth Abisado, Gabriel Avelino Sampedro · 2023
This research paper presents a novel approach to improving the longevity of IoT sensors by implementing a testbed architecture that leverages sensor data prediction. Unlike previous studies that primarily focused on enhancing prediction accuracy through increased hidden layers in Deep Learning (DL) models, this paper takes into account both prediction accuracy and energy efficiency as key metrics. To achieve this, a combination of DL models and the Fog-layer Data Prediction algorithm is deployed at the fog layer to forecast incoming data from edge devices. Various look-back intervals of historical data are evaluated to determine the most efficient approach. Simulation results indicate that LSTM algorithms with four and eight data look-backs achieved the highest accuracy rates compared to other DL models. Furthermore, the evaluation of the testbed architecture demonstrates that using sensor data prediction can enhance total energy efficiency by up to 20% compared to traditional IoT architectures.