Authors: Mitra Mousavi; Javad Rezazadeh; Omid Ameri Sianaki
Addresses: Islamic Azad University, North Tehran Branch, Tehran, Iran ' University of Technology Sydney (UTS), Sydney, Australia ' Victoria University Business School, City Flinders Campus, Australia
Abstract: Today, internet of things (IoT) has become an important paradigm. Everyday increasing number of IoT applications and services emerge. Smart devices connected by the IoT generate significant amounts of data. Analysis IoT sensor data using machine learning algorithms is a key to achieve useful information for prediction, classification, data association and data conceptualisation. Offloading input data to cloud servers leads to increased communication costs. Undertaking data analytics at the network edge using fog computing enables the rapid processing of incoming data for real-time response. In this paper, we examine the results of using different machine learning algorithms on fog nodes based on existing research. These results are low latency, high accuracy and low bandwidth. Also, this work presents the current fog computing architecture which consists of different layers that distribute computing, storage, control and networking and finally we investigate the challenges and open issues related to the deployment of machine learning on fog nodes.
Keywords: internet of things; IoT; fog computing; machine learning; fog-based machine learning.
International Journal of Web and Grid Services, 2021 Vol.17 No.4, pp.293 - 320
Received: 17 Nov 2020
Accepted: 17 Dec 2020
Published online: 06 Sep 2021 *