A survey on machine learning in Internet of Things: Algorithms, strategies, and applications

In the IoT and WSN era, large number of connected objects and sensing devices are dedicated to collect, transfer, and generate a huge amount of data for a wide variety of fields and applications. To effectively run these complex networks of connected objects, there are several challenges like topology changes, link failures, memory constraints, interoperability, network congestion, coverage, scalability, network management, security, and privacy to name a few. Thus, to overcome these challenges and exploiting them to support this technological outbreak would be one of the most crucial tasks of modern world. In the recent years, the development of Artificial Intelligence (AI) led to the emergence of Machine Learning (ML) which has become the key enabler to figure out solutions and learning models in an attempt to enhance the QoS parameters of IoT and WSNs. By learning from past experiences, ML techniques aim to resolve issues in the WSN and IoT's fields by building algorithmic models. In this paper, we are going to highlight the most fundamental concepts of ML categories and Algorithms. We start by providing a thorough overview of the WSN and IoT's technologies. We also discuss the vital role of ML techniques in driving up the evolution of these technologies. Then, as the key contribution of this paper, a new taxonomy of ML algorithms is provided. We also summarize the major applications and research challenges that leveraged ML techniques in the WSN and IoT. Eventually, we analyze the critical issues and list some future research directions. (C) 2020 Elsevier B.V. All rights reserved.