electronics

Review
Machine Learning-Enabled Internet of Things (IoT): Data,
Applications, and Industry Perspective
Jamal Bzai 1 , Furqan Alam 2, * , Arwa Dhafer 3 , Miroslav Bojović 1 , Saleh M. Altowaijri 4 , Imran Khan Niazi 5
and Rashid Mehmood 6

1 School of Electrical Engineering, University of Belgrade, 11000 Belgrade, Serbia

2 Data Science Department, SkilloVilla Private Limited, Bengaluru 560076, India
3 College of Computing and Informatics, Saudi Electronic University, Riyadh 13316, Saudi Arabia
4 Department of Information Systems, Faculty of Computing and Information Technology,
Northern Border University, Rafha 91911, Saudi Arabia
5 GMNG Entertainment Private Limited, Gurugram 122001, India
6 High-Performance Computing Center, King Abdulaziz University, Jeddah 21589, Saudi Arabia
* Correspondence: [email protected]

Abstract: Machine learning (ML) allows the Internet of Things (IoT) to gain hidden insights from the
treasure trove of sensed data and be truly ubiquitous without explicitly looking for knowledge and
data patterns. Without ML, IoT cannot withstand the future requirements of businesses, governments,
and individual users. The primary goal of IoT is to perceive what is happening in our surround-
ings and allow automation of decision-making through intelligent methods, which will mimic the
decisions made by humans. In this paper, we classify and discuss the literature on ML-enabled IoT
from three perspectives: data, application, and industry. We elaborate with dozens of cutting-edge
methods and applications through a review of around 300 published sources on how ML and IoT
work together to play a crucial role in making our environments smarter. We also discuss emerging
Citation: Bzai, J.; Alam, F.; Dhafer, A.; IoT trends, including the Internet of Behavior (IoB), pandemic management, connected autonomous
Bojović, M.; Altowaijri, S.M.; Niazi, vehicles, edge and fog computing, and lightweight deep learning. Further, we classify challenges to
I.K.; Mehmood, R. Machine IoT in four classes: technological, individual, business, and society. This paper will help exploit IoT
Learning-Enabled Internet of Things opportunities and challenges to make our societies more prosperous and sustainable.
(IoT): Data, Applications, and
Industry Perspective. Electronics 2022, Keywords: Internet of Things (IoT); outliers; data imputation; feature selection; machine learning; smart
11, 2676. https://doi.org/10.3390/ cities; smart homes; edge and fog computing; lightweight deep learning; Internet of Behavior (IoB)
electronics11172676

Academic Editor: Shinichi Yamagiwa

Received: 19 July 2022 1. Introduction

Accepted: 18 August 2022
The Internet of Things (IoT) is set to become one of the key technological developments
Published: 26 August 2022
of our times, provided we can realize its full potential. IoT is “a global infrastructure is
Publisher’s Note: MDPI stays neutral enabled using advanced services by interconnecting (physical and virtual) things based
with regard to jurisdictional claims in on existing and evolving interoperable information and communication technologies.”
published maps and institutional affil- IoT was named by the US National Intelligence Council (NIC) in a 2008 report among
iations.
the six vital civil technologies that could potentially affect US power. IoT is an enabler
of ubiquitous computing envisioned by Mark Weiser. IoT is no longer a technological
buzzword as described in [1], but a reality that links the physical world to the digital
world, revolutionizing how we look towards our surroundings. Currently, IoT is partially
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
implemented in bits and pieces due to a lack of availability of technology and other
This article is an open access article
constraints on the global scenario.
distributed under the terms and
1.1. IoT Industry and Market
conditions of the Creative Commons
Attribution (CC BY) license (https:// The IoT industry has attracted information technology giants like Microsoft, Cisco,
creativecommons.org/licenses/by/ Google, Amazon, Apple, and Samsung to invest in IoT-enabled hardware and software.
4.0/). According to the market research analytics companies Statista and Transforma Insights, the

Electronics 2022, 11, 2676. https://doi.org/10.3390/electronics11172676 https://www.mdpi.com/journal/electronics

Electronics 2022, 11, 2676 2 of 33

number of objects connected to the IoT is expected to reach 25–30 billion by 2030 due to the
massive influx of various IoT devices proliferating [2,3].
The primary purpose of these increasing numbers and types of IoT objects is to
produce valuable data about the entities present in the operating environment to make
smart decisions. This is achieved by providing access to the environment from which we
need information and analyzing past, present, and future data. These data allow optimal
decisions about us and our environments, possibly in real time. This massive, diverse
growth in the overall IoT landscape will produce $1–1.5 trillion in revenue annually. The
IoT landscape is illustrated in Figure 1. Although Europe is at the forefront in the early
adoption of IoT, South Korea tops the global ranking of connected things, whereas the USA
is far behind in this respect [4]. Figure 1 also depicts application areas of IoT: smart homes,
warning systems, smart shopping, smart gadgets, smart cities, intelligent roads, health care,
fire systems, threat-identification systems, tracking, and surveillance.

Figure 1. Infographic showing IoT landscape, stakeholders, and market forecasts.

1.2. Data Production

One of the significant functions of IoT is to provide technological infrastructure to
sense different activities and events happening in our surroundings. IoT is expected to
produce an enormous amount of data. These data would be created by various vendors,
giving rise to data as a service. For powering smart cities and societies to their full potential,
sharing and collaborating data and information will be the key to providing them with
sustainable and ubiquitous applications and services. The fusion of various types and
forms of data to enhance data quality and decision-making will be of prime importance in
ubiquitous environments. Data fusion is “the theory, techniques, and tools used to combine
sensors’ data, or data derived from sensory data, into a common representational format.” A
timely fusion and analysis of big data (volume, velocity, variety, and veracity) acquired from
IoT’s sensor networks enables accurate and reliable decision-making. However, managing
ubiquitous environments would be a grand challenge for IoT. Also, various sensors and
intelligent algorithms would play a critical role in addressing the above challenge.
Electronics 2022, 11, 2676 3 of 33

1.3. Machine Learning and IoT

IoT’s objectives are to understand what people want and how people think, predict
wanted and unwanted events, and learn to manage certain situations. For all of this, IoT
needs to understand the data produced by millions of objects. This understanding can
be gained by using machine-learning algorithms (MLAs). Machine learning (ML) in the
IoT paradigm can play a significant role that is imperative. IoT is ubiquitous by nature,
which means to be available anywhere is one of its primary goals [5]. ML will play a
significant role in this by digging out the data produced by thousands and millions of
connected devices. ML will add usefulness to IoT devices, and IoT can only be genuinely
ubiquitous [6]. Embedded intelligence (EI) will be at the core to enable IoT to play a
significant role in achieving its objectives. EI is the fusion of product and intelligence
to achieve better automation, efficiency, productivity, and connectivity [7,8]. Whether a
physical or virtual world, intelligence is acquired by learning.
The tendency of ML to find patterns may be the underpinning for human-like intel-
ligence. Further generalization of these patterns into more valuable insights and trends
provides an improved understanding of the world around us. The actual objective of ML
in IoT is to bring complete automation by enhancing learning that facilitates intelligence
through smarter objects [9]. ML gives IoT-enabled systems the potential to mimic human-
like decisions after training from the data and further improve their understanding of our
surroundings. The influence of information visualization on the human visual system is
enormous, making systems better understand data and insights [10]. Information visualiza-
tion brings several advantages to its users, like (1) better knowledge without much further
analysis of data and (2) using cognitive skills, and humans can better understand data. IoT
will replace several systems currently used, which are costly to implement and maintain,
with cheap sensor-based ML systems. For example, around 20,000 people lost their lives
in developing countries due to severe weather conditions. Mostly, weather monitoring
is done by radar-based weather-monitoring systems (WMSs). However, radar WMSs are
costly and unavailable in several parts of the world. An ML-enabled IoT system consisting
of a cheap sensor network that studies lighting and cloud patterns to predict the weather is
successfully deployed in economically backward countries like Guinea and Haiti [11].
IoT will not only impact how we see technology but also how technology can bring
progress and make our world more prosperous [12,13]. Every day, various aspects of our
lives are becoming easier and more connected through the IoT. ML brings intelligence and
pervasiveness to IoT. IoT can be an appropriate synonym for the word “heterogeneous,”
as it consists of various devices, network technologies, protocols, data types, applications,
and users. This heterogeneous nature of IoT brings several challenges to ML, which are:
• IoT will produce big data [14–16], but are all the data valid, do the data have biases,
and is it worthwhile to process all of it? These are some critical questions shaping the
reliability, accuracy, and efficiency of MLAs for the IoT domain.
• Not all IoT applications have ample data from which MLAs will learn quickly, but a
lot of small data are also produced for this new form of algorithms needed to learn
from scarce data [17,18].
• Sensing devices are not always accurate and reliable [19–24]. Outlier detection and
data imputation are some of the necessary tasks needed to be performed on data
before ML begins.
• The application area of IoT is enormous, as mentioned in Figure 1. Every application
has data with particular properties.
• At Google’s Zeitgeist 2011 event, Google’s chief scientist Peter Norvig famously said,
“We don’t have better algorithms than anyone else; we just have more data.” How-
ever, few researchers support the opposite. Which is better? A highly sophisticated
MLA [25], with more data [26–28] or limited but high-quality data [29,30]? This
question still has no answer and is one of the significant conflict areas among ML
researchers, as opinions vary.
Electronics 2022, 11, 2676 4 of 33

• Game-changer technologies such as IoT hold ample opportunities for businesses,

but also pose a high risk, and ML-enabled IoT could end up swallowing millions of
jobs [31,32].
Machine learning (ML) gives a brain to IoT-enabled systems to grasp the insight from
data produced by millions of IoT objects. In IoT, we see several MLAs learning from diverse
data, which makes ML completely different from IoT, as on the one hand, we will still have
traditional MLAs; on the other, we need a completely different set of MLAs. We will have
different classes of MLAs, and some will work based on simple, intuitive insights instead
of complex mathematical proofs [33].
Eric Brill and Michele Banko in early 2001 [34] published an interesting paper that
showed that more training data results in improved learning, rather than enhancing and
designing new MLAs. Big data will never be a problem in IoT. Billions of connected IoT
objects via the internet will produce massive data [35]. As a result, IoT-based ML consists of
algorithms that will learn from the colossal amount of data. The IoT domain (1) is partially
valid, as IoT is not all about big data, but also about small data [17,18]. Small data sets
contain minimal attributes. Small data can be used to describe the current state, trigger
events, and be produced by the aggregation of big data.
Governments, industries, and individuals have a broad spectrum of IoT-enabled
applications that take leverage from ML. Shanthamallu et al. [36] discussed MLAs and
their application areas in IoT. At the same time, Sharma and Nandal focused on “machine
learning as a service” (MLaaS), which is the fusion of ML and IoT infrastructure [37]. MLAs
can perform various tasks in IoT, as illustrated in Figure 2. Broadly, ML tasks can be seen
from IoT perspectives: (1) data quality and (2) pattern recognition. MLAs not only predict
from a treasure trove of data but also can be used to enhance data quality, ultimately
resulting in better learning. For example, MLAs are used to identify outliers and impute
data before training MLAs for prediction. In the proceeding sections, we are going to
discuss these in detail.

Figure 2. Machine-learning task in the Internet of Things (IoT).

1.4. Contributions
Most R&D endeavors on IoT have focused primarily on object and resource manage-
ment, object identification, access control, network, and connecting technologies. Instead of
focusing on major IoT R&D trends mentioned above, in this paper, we attempt to enhance
our understanding of how ML plays a critical role in shaping the IoT landscape. This
survey will work as an underpinning for IoT-based ML researchers. In Table 1, we have
given five major ML-enabled IoT surveys and their objectives. Our intention is not to
represent a comprehensive review of the literature. Still, this paper attempts to achieve the
more significant aim of enhancing ML’s understanding, usefulness, and significance for the
IoT domain. The main contributions of this work are fourfold:
Electronics 2022, 11, 2676 5 of 33

• Firstly, we classify IoT-related research and development work into three major per-
spectives (classes): data, application, and industry.
• Secondly, the paper gives insight into the current state-of-art research and develop-
ments in IoT with a specific focus on ML-related developments.
• Thirdly, the paper identifies emerging IoT trends that will use the machine at its core
to develop futuristic and sustainable solutions.
• Lastly, the paper helps the readers to identify future opportunities in IoT-based
ML research.

Table 1. Major machine learning-based IoT surveys.

No. Paper Year Objectives

1 Siow et al. [38] 2018 ML-based IoT survey of IoT analytics,

types, and infrastructure
Deep learning-based IoT survey of big
2 Mohammadi et al. [39] 2018
data and streaming analytics

3 Mahdavinejad et al. [40] 2018 Use case based on usage of machine

learning for smart-city environment
ML-based IoT survey of data
4 Alam et al. [41] 2017
fusion techniques

1.5. Paper Structure

The paper is divided into seven sections. As depicted in Figure 3, in Section 2, we
discuss IoT from a data perspective. In Section 3, we critically analyze the role of ML from
an application perspective, whereas in Section 4, we discuss IoT’s industry perspective.
Further, in Section 5, we discuss five emerging trends where the fusion of IoT with ML
will play a critical role. In Section 6, we classify challenges to IoT’s success in four classes:
technological, individual, business, and society. Finally, we conclude in Section 7.

Figure 3. The high-level structure of the paper.

Electronics 2022, 11, 2676 6 of 33

2. Data Perspective
Data add value to the IoT paradigm and are collected by using a variety of sensors, as
given in Table 2. IoT has both cheap and expensive sensors in its arsenal. For example, the
temperature detection sensor is cheaper than lidar, which is too costly. The type of sensor
used largely depends on the type of application of that data. Wild animal tracking sensors
will have lifelong battery life, as replacing batteries in wild animal tracking applications is
hard, whereas sensors like lidar, cameras, and radar continuously need a power supply to
function. Also, low-cost sensor data have such issues as outliers and missing values as their
hardware quality is limited. On the other hand, vision sensors bring many features, and
selecting only the best feature is challenging. In the proceeding subsections, we discuss IoT
data sources, data storage platforms, and the three types of data challenges with a specific
focus on machine learning.

Table 2. IoT paradigm concerning data sources, applications, and data challenges.

Type of Possible Data

Type of Data
Sensors Applications Challenges
Satellites, Autonomous Vehicles,
Vision Images & Videos Feature Selection
Robots, Tracking
Industry, Autonomous Vehicles,
Force, Pressure, Load Numeric Readings Outliers, Missing Values
Robots, Healthcare
Satellites, Autonomous Vehicles,
Motion & Velocity Numeric Readings Outliers, Missing Values
Robots, Tracking
Gas & Chemical Numeric Readings Pollution, Industry, Healthcare Outliers, Missing Values
Temperature, Fire warning, Weather,
Numeric Readings Outliers, Missing Values
Humidity, Moisture Industry, Healthcare
Satellites, Autonomous Vehicles,
Radar, Lidar, Proximity & IR Numeric Readings Outliers, Missing Values
Robots, Tracking, Defence, Space
Satellites, Autonomous Vehicles,
Numeric &
Positioning & Location Robots, Tracking, Outliers, Missing Values
Coordinates Readings
Mobile Applications
Level & Flow Numeric Readings Industry, Dams, Tanks Outliers, Missing Values
Sound & Vibration Numeric Readings Industry, Dams, Tanks, Healthcare Outliers, Missing Values

2.1. Data Sources

One of the major applications of IoT is sensing our surroundings and communicating
that data to the smart application, which will be used to predict and forecast using machine
learning algorithms. Further, the learning outcome is used to develop AI for making
decisions. Later, the decision is transformed into mechanical output using actuators [38].
Today, billions of devices with sensors surround our daily life. IoT produces and will
produce an enormous amount of data that needs to be stored, processed, and archived
for future needs. IoT infrastructures are not yet totally implemented, even in developed
economies. Developing economies like India, Malaysia, etc., are slowly working on mega-
smart-city projects that will use IoT infrastructure. An exciting work has been done by
Morais et al. [42] where they classify IoT data types into 19 common categories that are
in use. Also, they classify IoT sensor types too. We depicted in Table 2 the possible kinds
of sensors used in IoT, applications, and data challenges particular sensors bring. The
COVID-19 pandemic expedited the demand for IoT solutions.

2.2. Data Storage

IoT means an enormous amount of real-time data. For example, autonomous vehicles
alone can contribute to colossal amounts of data. For example, Wang et al. proposed
HydraSpace [43] multilayered storage architecture for storing autonomous vehicles’ data.
The cloud may be more flexible, scalable, and ubiquitous. However, it is not possible to
Electronics 2022, 11, 2676 7 of 33

have real-time data analytics from data stored on clouds. This makes edge- and fog-based
IoT data storage critical [44]. ML and AI available on edge devices can give real-time
insights from the sensed data. Later on, aggregated data can be stored in the cloud. The
transfer of data on edge first will create a more realistic and valuable IoT landscape. Some
popular and widely used IoT-based cloud storage services are AWS S3 (Amazon Web
Services, Seattle, WA, USA), IBM Watson IoT Platform (IBM, Armonk, NY, USA), Oracle
IoT Cloud Service (Oracle, Austin, TX, USA), and Microsoft Azure (Microsoft, Redmond,
WA, USA) [45,46].

2.3. Data Issues

Our increasingly connected world through IoT is a delicate blend of low-cost sensors
and distributed intelligence that will have a transformative impact on how we see the world.
This merger will produce more data than ever that hold valuable information. Sensing
data has critical quality issues, as sensing devices are not 100% reliable and accurate.
Preprocessing of IoT data is required before feeding it to MLAs to gain critical insights. As
depicted in Figure 4, three significant issues with sensed data, outliers, missing values, and
feature selection are discussed in the proceeding sections.

Figure 4. IoT-based data issues and solution landscape.

2.3.1. Outlier Detection

Outliers, also known as anomalies are the data patterns that differ from the rest of the
data and signify abnormal data behavior [47–49]. Outlier data observations are everyday
in highly dense sensor environments like IoT, due to: (1) low-cost sensors which mean
low quality, (2) weather conditions, (3) electronic inferences, and (4) data communication
errors [50–52]. Outliers must be detected rather than deleted or replaced by predicted
values, which is crucial to maintaining high data quality from which MLAs ultimately dig
out the key insights. Modern-day MLAs are not only used for gaining valuable knowledge
but also for improving data quality by detecting data aberrations [53]. Significant attention
has been given to outlier problems in wireless sensor networks (WSNs) [54–58], which can
also be seen as a subset of IoT.
Several critical surveys exist that primarily focus on addressing the problem of outliers
in the IoT landscape. Alghanmi et al. [59] did a general-purpose comprehensive study on
Electronics 2022, 11, 2676 8 of 33

ML-powered anomaly detection and discussed available IoT datasets used for this purpose.
On the other hand Cook et al. [60] examined how to detect outliers in IoT-based times series
data. In contrast, Diro et al. [61] saw outliers’ detection as a way to make IoT networks
more secure. Further, a more recent survey by Samara et al. [62] focused on statistics-based,
clustering-based, nearest neighbor-based, classification-based, artificial intelligent-based,
spectral decomposition-based, and hybrid-based for outlier detection. Commonly used
outlier detection (OD) approaches are based on statistics, distance matrix supervised, and
unsupervised ML. One such MLA is SVM, which has an explicit mechanism to handle
outliers robustly [57,63]. Resource overhead is one major issue with SVM-based OD.
An unsupervised centered quarter-sphere SVM with low computational complexity and
memory usage for online OD is proposed in [64] which outperforms previous offline OD
methods based on SVM [65].
In unsupervised learning, k-means are a simple yet popular choice along with hi-
erarchical clustering for OD, as critically analyzed by Garcia-Font et al. [58]. Similarly,
Münz et al. [66] focused on k-means clustering for OD in traffic data. One major drawback
with k-means is that it computes a set of k centers to reduce the sum of squared distance.
Multiple works [67,68] show that there are two issues related to it: (1) outliers can pull
these centers, and (2) rather than rejecting outliers, it can be possible for outliers to form
their cluster. As a solution for this, a robust version of k-means is proposed by Statman
et al. [69] known as k-means+++. Another classifier quite popular for OD problems is naïve
Bayes because of its ease of use and simplicity [70,71]. Parto et al. [72] evaluated classical
Bayesian techniques with slight modification for OD in streaming IoT platforms in the
manufacturing industry. Further, similar to [66], Lam et al. [73] addressed the OD issue in
traffic data using a fusion of naïve Bayes and Gaussian mixture-model techniques.
C4.5 and its successor C5.0 MLAs are highly accurate and efficient modern-day classi-
fiers that outperform the best in the business classifiers, as analyzed in [74]. However, little
attention has been given to C4.5 and C5.0 for OD, particularly in the IoT environment, as
they have high precision, minimum memory usage, and fast processing. Today in every
field, we are witnessing the increasing use of deep-learning algorithms due to their ability
to provide highly accurate predictions and forecasting output. These algorithms can under-
stand highly complex datasets, which gives them an edge over others. Luo et al. proposed a
distributed outlier detection method for sensor networks that uses deep autoencoders [75].
The technique can produce a high detection rate with minimum communication overhead,
which is necessary for IoT-based sensor networks. Similarly, Diro and Chilamkurti [76]
used deep learning for cybersecurity purposes. Their work shows that the deep model is
more capable of detecting anomalies than shallow learning.
In IoT, MLAs for OD can be divided into three classes. First-class algorithms execute
offline and online. The second class of IoT algorithms will come from where intelligence
and data lie. Finally, the third class of OD algorithms is based on the accuracy requirement
of IoT applications. A detailed illustration is given in Figure 5.
Electronics 2022, 11, 2676 9 of 33

Figure 5. Classification of future outlier detection algorithm that adopt machine learning for
IoT application.

2.3.2. Data Imputation

The IoT ecosystem relies heavily on hardware like sensors and RFIDs for sensing data.
Sensors are not reliable, an established fact [50,51,77]. One of these outcomes is missing
values produced in IoT-based applications. The missing-values problem arises due to
various reasons like synchronization problems, unstable wireless communications, sensor
failure, power loss, and weather conditions [78]. Two techniques used to handle missing
values in IoT data are (1) deleting the missing data instances and (2) replacing the missing
value with predicted data, a process known as data imputation [79]. Much attention has
been given to developing data imputation algorithms in several areas, such as natural
sciences, census surveys, WSN, robotics, and scientific applications.
ML algorithms are widely used to impute the missing values. A lazy learner knows,
as the KNN algorithm is a nonparametric method. It is straightforward to implement
and simple to understand MLA. KNN is one of the top MLAs [80], and data imputation
algorithms based on KNN are widely used [81–84]. More complex and computationally
extensive supervised learning algorithms than KNN, such as SVM, are also widely used for
data imputation and handle both linear and nonlinear data efficiently [80,85]. In various
papers, SVM is used multiple ways to deal with missing values [86–89]. However, SVM may
not be too accurate with more than two class problems. A slightly younger supervised MLA
than SVM, known as random forest (RF), based on ensemble learning, was introduced by
Leo Breiman and Adele Cutler [90]. RF-based DI is widely used in practice. Such methods
based on RF MLA are presented in [91,92] that use proximity from RF to impute missing
data values.
A more advanced, memory-efficient, and fast MLA, C4.5 is considered one the best
MLAs [80], and Jerzy et al. [93] concluded that C4.5 is one of the handiest algorithms
for dealing with missing data. C4.5 uses an internal data imputation mechanism based
on a probabilistic approach [94]. A few works also indicate that rather than using the
C4.5 internal data imputation mechanism, KNN-based DI for C4.5 results in improved
prediction accuracy [94,95], not just supervised machine learning, and the DI problem
is unfolded from an unsupervised machine learning perspective. Several novels and
hybrid data imputation algorithms are proposed, such as K-means [96–101], fuzzy c-means
with support vector regression [102], fuzzy clustering [103], feature selection, and cluster
analysis [104], and multiple imputation using gray-system theory and entropy based on
clustering (MIGEC) [105]. Another interesting algorithm is based on ANNs, which mimic
the neural system of the human brain. ANNs are incredibly efficient in data imputation.
Various novel and hybrid ANNs for data imputation are proposed, such as fuzzy min-max
neural networks [106], particle swarm optimization (PSO), evolving clustering method
(ECM), and auto associative extreme learning machine (AAELM) [107], ANNs and case-
based reasoning (CBR) [108], general regression and auto associative ANN [109], and ANN-
Electronics 2022, 11, 2676 10 of 33

based emergent self-organizing maps [110]. Except for C4.50 s MLA, dealing with missing
values can be expensive in terms of storage and/or prediction-time computation [111]. The
fundamentals for imputing missing values will remain the same in IoT as in other domains.
However, we envisage that data imputation will move more towards real-time processing
of missing values in context with IoT’s future scope, particularly for IoT applications.

2.3.3. Feature Selection

The problem of identifying the most critical information, potentially overpowering
the amount of data, has become increasingly significant. High-dimension data introduce
several problems for MLAs which are: (1) may reduce accuracy, (2) high computation,
(3) increased memory requirements, and (4) visualization becomes tough. Selecting the
most relevant feature subset is known as feature selection (FS). There are three FS strategies:
(1) wrapper-based FS, (2) filter-based FS, and (3) embedded FS. Figure 6 shows the categories
of feature-selection methods. Several studies have been done to study the effect of FS on
ML methods [112–116], which proved that FS improves the accuracy of various ML models
and decreases computational cost.

Figure 6. Various categories of feature-selection methods.

In IoT, particularly in highly dense sensor setups [117–119], data produced are massive
and often possess high dimensions. Therefore, FS methods must be exploited to gain
more accurate information. Several extensive surveys of various feature-selection and
dimensionality-reduction approaches can be found in the literature [120–122]. MLAs such
as KNN [123–126], SVM [127–129], decision tree [130,131], AdaBoost [132–134] random
forest [135–138], naïve Bayes [139,140], regularization [141–143] and relief-F [144–147],
entropy evaluation criteria [148] are extensively used for identifying the most relevant
variables in the datasets.
Guo et al. concluded in [7,8] that IoT will ultimately EI-enable IoT, which will serve
this goal. The FS method will also be helpful in terms of data reduction, apart from its
other advantages [112–116]. For example, EI-enabled smartphones and home appliances
will not have much processing power, storage needs are limited, and data are relevant only
to predict an event. Most of the ongoing research on FS is based on offline FSA methods
that can be suitable for most applications in domains like natural sciences and geography.
However, for IoT, online FSAs are required for most of its applications. As in the IoT
ecosystem, scenarios will change quickly, and most of the decision-making will be based
on streaming data.
ML will significantly address the above data issues related to outliers, missing values,
and feature selection. Most IoT sensors will be low-cost hardware that tends to have a
Electronics 2022, 11, 2676 11 of 33

temporary malfunction. In the proceeding section, we discuss IoT applications that will
use the processed data produced by sensors.

3. Applications Perspective
IoT has evolved beyond what Atzori et al. [1], defined it. Today, it is seen as a discovery
that has the potential to change the world in the same way as electricity did to humankind.
Xu et al. systematically provide a concise view of current IoT application areas, R&D trends,
and challenges for IoT in industries to provide an understanding of IoT developments in
industries. Data for IoT are as important as electrons are for electricity. In this section, we
examine ML developments in IoT and classify IoT applications according to [149,150].
According to a United Nations report, more than half the world’s population lives in
cities, due to better jobs, education, health care, and living conditions [151], putting extraor-
dinary pressure on municipalities, urban development departments, and governments to
provide sufficient resources. Due to this fact, the “smart city” concept has recently drawn
significant attention from governments worldwide, especially in developed [152,153] and
developing economies [154,155]. Smart cities are now an essential part of urban devel-
opment planning. There exists no formal definition of a smart city. However, it can be
defined as the product of accelerated development and advanced information technology,
which aims to improve citizens’ socioeconomic conditions and enhance the overall quality
of living.
IoT is about connecting physical devices using the internet to facilitate the smooth
exchange of information. The smart city dream would not be possible without the technical
support of IoT, which is inevitable to achieve smart city aims—Zanella et al. termed
it “urban IoT” [156]. In the background of urban IoT, an immense amount of data are
produced by “things.” Gaining key insights from these data is a critical problem that ML
can solve. ML in urban IoT is a bit different from other domains, due to its heterogeneous
nature in terms of devices, data, and applications, as seen in Figure 7.

Figure 7. Smart city landscape.

3.1. Smart Grids

Smart grids enhance energy availability and efficiency, providing uninterrupted power
supply to cities, towns, and businesses by minimizing power wastage, reducing faults, and
optimizing power supply to cope with high energy demands [157,158]. Power-grid failures
are rare, but they result in the loss of millions of dollars, blackouts, and social ataxia. Smart
grids supply power in a more distributed, adaptive manner. Pinning hopes on smart grids
for better power management is achievable through IoT. According to Randal Bryant et al.,
Electronics 2022, 11, 2676 12 of 33

the contribution of ML to the success of smart grids will be enormous and beyond what we
see today. It describes the energy-space domain where MLAs are expediting the progress
of the “data to knowledge to action” paradigm [159].
Further, Zhang et al. [160] critically examine smart grids’ potential applications of
deep learning, reinforcement learning, and integration. However, IoT-based smart grids
are also bringing security challenges. With the availability of treasure-trove data and
MLAs [161–166], we can find critical power usage patterns and consumer preferences.
This will maximize the reliability of power grids and further share essential insights with
consumers and power companies to improve or design better power infrastructure for
future challenges.
Electricity-demand forecasting (EDF) has gained significant attention, and it is a critical
task in strategic planning for power companies. EDF impacts the operational decisions
in smart grids, as pointed out in [167]. MLAs are the primary tools for EDF, which learn
from data and predict. In conventional power grids, EDF is based on historical power
consumption data. However, smart grids are the end product of the merger of IoT and
power grids. As a result, more diverse data are available from various IoT applications, as
mentioned in Figure 2, which can be used for highly accurate predictions.
In [166,168,169], the authors examined some of the widely used MLAs based on their
effectiveness in facilitating operational decisions in smart grids. Nonlinear MLAs like
ANNs and SVM are most persuasive for EDF. ANNs are very potent for modeling any
nonlinear relationships and complex behaviors of smart grids. Various types of ANNs
are used for demand forecasting: BP [170,171], radial basis function (RBF) [172–174], mul-
tilayer perceptron (MLP) [175], and optimized and hybrid ANNs [170,176–178]. EDF by
SVM [179–182] can handle noise better with minimum overfitting. ANNs and SVMs are
highly accurate with conventional EDF. However, a novel deep-learning model known as
factored conditional restricted Boltzmann machine (FCRBM) for EDF shows significant
improvement in prediction accuracy [183]. ANNs, SVMs, and FCRBM are computation-
ally expensive for IoT to envision. EDF depends not on conventional grid data, but on
several evolving factors in IoT ecosystems like weather, social events, individual prefer-
ences, power-grid performance, and maintenance. Smart grids will be essential for human
urbanization prospects; nevertheless, their management is critical. This is achievable by
ML-enabled IoT. However, this also brings some challenges related to the security of power
grids connected by IoT infrastructure [184].

3.2. Smart Traffic and Transportation

The value IoT brings to all the traffic solutions to their customers through its smart,
connected “things” is beyond what we have seen today. Urban mobility is the critical
application of smart traffic and transportation solutions, as it also enhances the chances
of accessibility of other services to the people. Throughout the world, cities are getting
bigger. The challenging issues for cities include traffic congestion, increased pollution,
and economic losses caused by traffic delays and road accidents. ML-enabled IoT strategy
provides the opportunity for creating value from connected data, including better services
and accelerated innovation [185,186].
In developed countries, the road infrastructure is highly advanced and well main-
tained. Contrary to this, road infrastructure suffers from maintenance issues in developing
economies. Roadway surface disruptions and obstacles (RSDOs) are widespread, resulting
in accidents, driving problems, traveling, and transportation delays. Gónzalez et al. in [187]
used acceleration sensing data to classify patterns related to speed bumps, potholes, metal
humps, and rough roads by using logistic regression and ANN MLAs. Another work [188]
that addresses the same issue identifies RSDOs by using a combination of supervised and
unsupervised ML with the help of data collected by the Street Bump smartphone appli-
cation. Traffic monitoring is critical in controlling traffic congestion, which is achieved by
identifying traffic patterns by analyzing vehicle movements using a granular classification
in [189] and regression analysis in [190]. Other applications of ML are intelligent traffic light
Electronics 2022, 11, 2676 13 of 33

management, which was achieved using Q-learning [191], and ANNs and reinforcement
learning [192].
Autonomous vehicles (AVs) are another area that will revolutionize the transportation
industry. AVs depend entirely on ML, eventually developing AI to drive without human
interference. ML algorithms are used for tracking and identifying moving and stationary
objects. Alam et al. [193] proposed a method to recognize objects in the driving scene
by integrating deep learning and decision fusion. Tesla and Google [194,195] are some
technological titans utilizing ANN and DL in their AVs to detect objects in the driving scene.

3.3. Smart Homes

IoT-enabled smart homes (SHs) are a technology concept that facilitates the complete
automation of operations of household devices and home appliances via the internet.
Context awareness is an important aspect of smart homes, as it improves the comfort
and safety of users. However, direct interaction between the user and the environment
decreases. The MavHome (Managing an Intelligent Versatile Home) project uses the
coupling of multiagent systems and probabilistic MLA for making home environment
response a rational agent [196] to maximize inhabitant comfort and minimize operating cost.
A more advanced context-aware model uses back-propagation ANN for service selection
and a temporal differential class of reinforcement learning algorithm for adaptive context
awareness, as user preferences do not remain the same over time. The main advantage
of [197] over [196] is that no predefined model is required for the context-aware system.
Modeling is automatically done based on the user’s feedback on the service.
SHs can make rational decisions for automation. This is achieved by tracking and
predicting the inhabitant’s mobility patterns and usage of devices. The active LeZi pre-
diction algorithm based on the Markov chain is proposed in [198], which can understand
subsequent event patterns. An important area that gained a lot of attention in enhancing
the automation of SHs recently is human activity recognition (HAR). Human behavior
prediction by activity recognition is made in [199] using algorithms based on deep learning.
Some comparative analysis of various ML algorithms exists, which showed their perfor-
mance with IoT-based HAR data. Fahad et al. compared the accuracy of five MLAs for
correctly recognizing smart home activities. SVM and evidence-theoretic KNN showed
higher accuracy than the probabilistic ANN, KNN, and NB in HAR [200]. In contrast, Alam
et al. [74] compared eight ML algorithms and concluded that DL performance in terms
of prediction accuracy is the best. Taiwo et al. [201] proposed a deep-learning model for
motion classification using movement patterns that is used to improve power usage in
homes. However, the DL algorithm is computationally expensive. Also, other work [74]
highlights that the C5.0 algorithm performs very close to the DL algorithm.
HAR is divided into two parts. Firstly, clustering of activity patterns, and secondly,
activity-type decisions. However, many related kinds of literature focus on one part only,
which results in performance degradation. An unsupervised MLA K-pattern is used to
classify complex user activities to answer this issue. Then, ANN is used to train and predict
user activities [131]. K-pattern MLA shows improved accuracy for high-volume IoT data
in terms of temporal complexity and cluster-set flexibility. HAR gives more control and
automation to smart homes. Better power optimization can be achieved by switching on/off
lights, fans, and home appliances. Emergency health conditions can also be identified, and
alerting others can avoid loss of life.

3.4. Smart Health Care

IoT is revolutionizing the health-care industry by bringing up new and advanced
sensors connected to the internet, producing essential data in real time. Islam et al. com-
prehensively explained IoT in health care, platforms, application, and industry trends
for smart health-care solutions [202]. The objectives of smart health-care applications are:
(1) improved and easy access to care, (2) increased health-care quality, and (3) reduced
health-care costs. The key to achieving the above objectives is to perceive patterns and
Electronics 2022, 11, 2676 14 of 33

critical insights from health-care data [203,204]. Automated assessment of individual well-
being and alerting others to any health risk for the patient is a widely researched topic.
In [205], an intelligent system is developed to monitor the well-being of individuals in
their home environments. An ML-based method is used to automatically predict activity
quality and automatically assess cognitive health based on activity quality. SVM, principal
component analysis (PCA), and logistic regression MLAs are used to quantify activities and
further predict cognitive health. Dawadi et al. also address automated cognitive health as-
sessment using ML. Supervised and unsupervised ML scoring models are used to quantify
and determine boundaries between activity performance classes and cognitive assessments
performed [206]. Cognitive systems can understand, reason, and learn, helping to spur
discovery and decreasing the effort required to populate research studies effectively.
Further, in [207,208], solutions for physiological monitoring, weight management, and
cardiovascular disease monitoring are proposed. In [207], a wearable armband multisensor
system known as BodyMedia FIT performs constant physiological tracking and weight
management by exploiting ML. The system has been commercially available since 2001 and
uses regression analysis to classify activities. In [208], the mobile machine learning model
for monitoring cardiovascular disease (M4CVD) is proposed. It uses mobiles to monitor
heart diseases. M4CVD locally analyzes trends of vital health signs by contextualizing
them with clinical data sets. SVM is used to examine features extracted from clinical data
sets and wearable sensors to classify a patient as at risk or at no risk of cardiovascular
disease, and has shown high accuracy in identifying patients at risk [208]. IBM Watson
provides a large-scale IoT-enabled cognitive health-care solution that covers a broader
spectrum of patients. It combines the power of health-care data with MLAs to give new
insights [209]. ML-enabled IoT health-care solutions enhance individuals’ proactive and
preventive health-care interventions and reduce health-care costs, whereas cognitive care
provides modern mechanisms for health-care specialists to connect with their patients,
improving diagnostic certainty and reducing error rates. IoT-based health-care solutions
can help in finding insights that can help raise the quality of health care across the globe.

3.5. Smart Supply Chain and Logistics

We are seeing many IoT applications in industries that are evolving and growing
daily. IoT produces enormous amounts of data coupled with the latest communications
technologies. Real-time data analytics helps businesses meet consumers’ demands in
today’s developing economies. Supply-chain management (SCM) epitomizes the impact
of IoT in the manufacturing industry. Ellis et al. explained how IoT-enabled analytic
applications would revolutionize SCM [210]. Some of the immediate benefits of IoT in
SCM, as highlighted by Barun [211], are:
• IoT “things” can communicate promptly, allowing the possibility of knowing where
they are at all times.
• Object tracking facilitated by IoT results in improved asset and fleet management,
which means well-planned scheduling, better routing, and on-time product deliveries.
• Better control of mobile assets with IoT means knowing where they are and how they
are used.
• Downtime will be audited closely in real time.
• It increases logistics transparency.
All these benefits are brought together in the broader scenario to make SCM more
efficient and sophisticated.
IoT means more data, more connected “things,” and a high degree of automation.
With many entities, such as vehicles, shipping containers, packages, and return shipments
as the origin of data, businesses require more advanced and sophisticated methods to
ingest and critically scrutinize IoT data. ML-enabled IoT gives SCM automated “sense,
decide, and reply” capabilities [212]. One of the crucial determinants of effective SCM is the
ability to recognize customer-demand patterns and react accordingly to the changes in the
face of intense competition. MLAs have shown promising results in demand forecasting.
Electronics 2022, 11, 2676 15 of 33

For demand forecasting, MLAs like ANN, recurrent ANN, SVM, NB, and linear regression
are compared, and SVM produces highly accurate forecasts [213]. ML-enabled IoT can
significantly enhance the efficiency of logistics and SCM efficiency. Zhengxia et al. proposed
an advanced logistics monitoring system based on IoT. It has various functions to support
the argument of multiple services in one place. One of the essential services is data
acquisition and processing, which shows that its data analysis and forecasting show MLAs
in modern logistics are a must-have [214].
Fraudulent imitation of packaging and products is known as counterfeit. It is a severe
problem for global supply distribution chains. As a solution for this, an anticounterfeit
deterministic prediction model (ADPM) is proposed in [215]. ADPM identifies counterfeit
by the Monte Carlo (MC) MLA. ADPM examines the product attributes by analyzing and
calculating the correlation coefficients among objective features. In other literature [216],
the authors tried to apply a machine learning-based approach with statistical techniques
to detect counterfeits. In this section, we review how IoT, ML, and the manufacturing
industry can join together to take on the challenges presently faced and streamline industry
processes with automation. Suppose all the discrete processes that used to take place in
silos can be observed and managed through the analysis of the data provided to MLAs. In
that case, the holy grail of proper supply-chain optimization may be within reach.

3.6. Smart Social Applications

Apart from the technological aspect, IoT can affect social aspects of human life more
than we can imagine. ML-enabled IoT can be used to find the public’s mood on a particular
issue and discover a pattern in social application data for event exploration. With the help
of connected devices like smartphones and tablets, opinions can be formed and public
perceptions can be analyzed by exploiting ML.
Opinions are the core of almost all human activities and are key influencers of our
behaviors. Our beliefs and perceptions of reality and the choices we make are, to a consid-
erable degree, conditioned upon how others see and evaluate the world. For this reason,
when we need to make a decision, we often seek out the opinions of others. This is not
only true for individuals but also for businesses. In [217], Liu gives an in-depth intro-
duction to this fascinating problem and presents a comprehensive survey of all possible
methods, including ML, that can be potential candidates, in addition to the latest devel-
opments in the field. Opinions can be predicted by analyzing public sentiment. In [218],
the authors proposed a sentiment-analysis technique that can translate the sentimental
orientation of Arabic Twitter posts based on novel data representation and MLAs. The
proposed approach applied many features: lexical, surface-form, syntactic, etc. We also
used lexicon features inferred from two Arabic sentiment word lexicons. The authors used
several standard classification methods to build a supervised sentiment-analysis system
(SVM, KNN, NB, DT, and random forest). Similarly, to [218,219] supervised classification
algorithms, such as SVM, KNN, and NB, are used for Arabic sentiment analysis, whereas
in [220] domain-specific sentiment analysis is done using MLA. Also, these days social
media analysis can be used to identify threats and unwanted events, as in [221], MLAs are
used for feature selection, and then only relevant text in the tweets is classified using SVM,
NB, and AdaBoost MLAs.
Similarly, complex events were identified in [222] using adaptive moving window
regression (AMWR) for dynamic IoT data streams. The emergence of ML in IoT gives us
three main advantages: (1) we are more connected, (2) more informed, and (3) actions can
be highly automated. ML makes IoT able to think and decide. The coupling of these two
gives us the power to sense, analyze, and predict the events in our social environment.

3.7. Smart Environment Control

One of the primary goals of IoT and smart cities is to make our societies more pros-
perous. Prosperity cannot be achieved until or unless cities provide a healthy living
environment to their residents. Clean water and good-quality air are significant issues for
Electronics 2022, 11, 2676 16 of 33

more than half the world’s population. IoT with smart applications can greatly change
this scenario. For example, IoT-based applications, such as eWater and sustainable-water-
management applications, are used to provide clean water in Gambia [223]. The world
has less clean drinking water because of manmade water pollution. The first step in
reducing and managing water pollution is identifying where water is polluted and by
how much. Shafi et al. [224] proposed a water-pollution detection method based on deep
neural networks.
Similarly, Mishra [225] proposed an IoT-based air-quality monitoring system. Several
machine-learning algorithms, such as linear regression, random forest, and XGBoost,
are used for forecasting and prediction. The model can be deployed for real-world use.
Similarly, Elvitigala and Sudantha in [226] used linear regression to compute pollutant-level
gases. Smart cities can leverage the fusion of IoT and machine learning to enhance the
automation of water, land, and air-pollution management operations to provide a safer,
healthy living environment that will ultimately result in a prosperous society.

3.8. Emergency and Disaster Management

Deployment of IoT infrastructure can significantly enhance our capacity to speed up
relief efforts during any emergency or disaster. Ray et al. [227] comprehensively examined
the IoT paradigm from its application area to data analytics based on machine learning
with a specific focus on disaster management. Forest fire is one of the areas where a
prompt forest-event prediction is an important application that can take from leverage
IoT infrastructure. For example, reaction time must be significantly less in the event of a
forest fire, as they propagate very quickly. However, this IoT-based system can predict the
wrong event due to outliers. To deal with such a problem, Nesa et al. [228] proposed an IoT
architecture that detects the data errors and events in an IoT-based forest environment using
classification and regression trees (CARTs), random forest (RF), gradient boosting machine
(GBM), and linear discriminant analysis (LDA). RF outperformed the other three classifiers.
Similarly, Salehi and Rashidi [229] categorized existing unsupervised machine-learning
methods for detecting outliers for the real-world application of forest fire prediction. We
witnessed during the last decade the destruction caused by a tsunami in terms of loss of
life and property. IoT infrastructure and machine learning can play a significant role in
developing an effective system to warn people of an expected tsunami. Pughazhendhi
et al. [230] addressed this issue by developing a tsunami early warning system. A tsunami
is predicted from earthquake data by an RF classifier. Further work [231] explained in
detail how Japan’s tsunami system works, one of the best in the world today.

3.9. Smart Security and Access Control

In the IoT environment, sensitive data are collected and transferred to their application
with partial or no human interference, which gives rise to the challenge of protecting the
security and privacy of millions and billions of users. There should be techniques for
access control to limit access to these data and information [232]. Attacks like denial of
service (DoS) attacks and distributed DOS, spoofing attacks, jamming, and eavesdropping
are prevalent and real threats to the security and privacy of user data and applications
in the IoT environment. Xiao et al. [233] proposed various classification, clustering, and
reinforcement learning-based access-control methods with the larger aim of protecting
overall user privacy in the IoT environment. In a more recent work [234], Hussain et al.
systematically reviewed different attacks, current state-of-the-art solutions, and challenges
for security in the IoT paradigm. Several critical security gaps were discussed. The authors
proposed extending machine-learning and deep-learning techniques, which are confined
to developing intelligence, as a security solution in the IoT paradigm. Also, [234] gave
future directions for ML- and DL-based access-control solutions. In one such work [235],
various types of attacks and anomalies in the IoT environment are predicted using several
ML algorithms, such as support vector machine (SVM), logistic regression (LR), random
forest (RF), decision tree (DT), and an artificial neural network (ANN). Decision trees and
Electronics 2022, 11, 2676 17 of 33

ANN performed better than the others did. Similarly, Khalifa et al. [236] critically examined
several biometric-based access-control methods used for feature extraction and classifi-
cation, such as Fisher discriminant analysis, linear discriminant analysis, learning vector
quantization, and ANNs. In addition, their advantages and disadvantages were discussed.
In [237], a deep learning-based method was introduced for smart-home application to limit
the access of pets and humans to consumer appliances. Interestingly, the proposed method
in [237] uses limited computing resources.

4. Industry Perspective
Predominantly, IoT remains at the initial stages of development and adoption by
the information technology industry (ITI). Slowly but steadily, the future worth of IoT is
envisioned by ITI. Driven by hopes, market trends, and statistics, a lot of R&D is going
on. ITI giants like Cisco, Microsoft, Google, IBM, Oracle, and SAP are at the forefront of
making our environment smarter by designing new IoT-enabled software platforms and
hardware. Increasing the use of IoT infrastructure will significantly enhance and speed up
the adoption of Industry 4.0, which will revolutionize industry practices [238].
Digging out key insights, or in simpler words making sense of IoT-generated data, is
one of the biggest problems in IoT. ML can tackle these issues. Another significant problem
is bringing ML to the masses, apart from the economic worth that IoT holds. In light of
these critical facts, ITI starts by adding MLAs as they collect more data for their IoT-based
systems. Some popular IoT-enabled ML systems are IBM Watson, Google TensorFlow,
Microsoft Azure, and Splunk, which are discussed here.
Microsoft Azure is a cloud computing platform created by Microsoft [239,240]. Joseph
Sirosh, corporate vice president of ML at Microsoft, says, “Every day, IoT is fueling vast
amounts of data from millions of endpoints streaming at high velocity in the cloud . . .
in this new and fast-moving world of cloud and devices, businesses can no longer wait
months or weeks for insights generated from data.” The reflection of his comments is
quite evident in Azure’s recent developments. The Azure cloud platform added ML with
advanced analytics to expand big data capabilities and be ready to tackle IoT. Services
such as Stream Analytics and Azure Event Hubs are intended to help customers process
data from devices and sensors in the IoT ecosystem. Scott Hanselman, principal program
manager for Microsoft Azure, demonstrated how this platform integrates several things
and facilitates ML for IoT [241].
Another interesting development came from IBM in the Watson software platform [242,243]
initially developed for answering in the quiz show Jeopardy. IBM Watson is a technology
platform that uses natural language processing and ML to disclose insights from vast
amounts of unstructured data. IBM Watson is more about cognitive IoT computing [244].
For example, a car owner wants to ask about the predictive maintenance date of a particular
auto part. Watson achieves this by analyzing machinery performance and breakdown
time with the help of sensor data gathered over time. Figure 8 illustrates the steps of
IBM Watson.
Watson APIs for IoT help to accelerate the development of cognitive IoT solutions and
services on the IBM Watson IoT Platform. By using these ML-enabled APIs, you will be
able to build cognitive applications that:
• enable a high degree of interaction with humans with the help of text and voice
• perceive images and scenes
• perform ML from sensory inputs
• establish data correlations with external data sources, such as weather or Twitter.
Guo et al. [7,8] presented ongoing efforts toward EI for smarter objects. Their work
also highlights the future transition of today’s IoT to EI-enabled IoT. The importance of
their work can be seen in the recent announcements of global collaboration by IBM Watson
and Cisco for combining the power of Watson IoT with edge analytics [245,246]. This
development also shows IBM’s willingness to scale down the unnecessary data transfer to
the cloud using edge analytics. Cisco’s fog-computing endeavors will be highly valuable in
Electronics 2022, 11, 2676 18 of 33

distributing intelligence at the edge. Watson’s role in this partnership is to provide a small
piece of code, informing the software of the exciting data for a particular requirement.

Figure 8. Step of how IBM Watson finds critical insights.

An exciting development came from search giant Google in the form of TensorFlow
(TF), an open-source ML platform [247]. Several Google products are now using TF. For
example, Google Photos, Gmail, Google search, and speech recognition utilize TF. A signifi-
cant advantage of TF is that it is highly scalable and can run on several systems, servers,
personal computers, smartphones, and other mobile devices. Users can execute custom-
distributed MLAs. The potential of deep learning can be exploited by using TF. Like
Microsoft and Google, another US-based multinational corporation, Splunk, introduced
IoT-enabled ML software known as Splunk (product), which is excellent in gaining fun-
damental insights from operational data. It handles big data efficiently and augments
maintenance and fault diagnosis from IoT-generated data [248]. It consists of around
300 new MLAs [249]. Splunk stresses the fact that its ML system will benefit nontechnical
users. Splunk integrates with popular IoT platforms and services that can be seen as a
boost for the broader acceptance of Splunk.
How important ML is becoming for future IoT is quite evident in the latest develop-
ments of Amazon Web Services (AWS) IoT, which in early 2016 integrated with Amazon
Machine Learning (AML) [250]. As Google, IBM, and Microsoft offered cloud-based
machine-learning platforms, Amazon has been obliged to step up with its product to meet
market demand. AWS and AML integration allows users to create ML models without
knowing much about ML. However, the AML platform offers an easy way to do simple
data analytics, but this also confined it within a boundary [251]. Several other companies
are in the market, offering application-specific ML solutions for IoT. Recently, market
research company CB Insights used the Mosaic algorithm to classify promising start-using
ML and DL algorithms to provide predictive insights from IoT-generated data [252]. The
application area of ML in IoT is enormous. Undoubtedly, the challenges and opportunities
presented by IoT [4,6,7,19,22,188] are driving the growing interest in ITI in developing
ML-enabled IoT.

5. Emerging Trends
IoT has had some interesting emerging trends in the last few years, such as edge
computing, fog computing, deep learning, and connected autonomous vehicles. Also, in
the previous few months, we have seen IoT used successfully in managing and controlling
the COVID-19 pandemic. All the above mentioned emerging trends are discussed in the
proceeding subsections.
Electronics 2022, 11, 2676 19 of 33

5.1. Internet of Behavior (IoB)

IoT is a fusion of sensors, actuators, and connectivity technologies, whereas the Inter-
net of Behavior (IoB) is a fusion of IoT, intelligence, and behavioral science. IoB can be seen
as an extension of IoT. Its goal is to better understand data that will facilitate better product
development and promotion, focusing more on evolving human psychology [253]. Javaid
et al. [254] posited that the inception of IoB can change the dynamics of product or service
design, marketing, and customer services due to its ability to understand and modify con-
sumer behaviors based on their comportment, tastes, and imaginations. Pinochet et al. [255]
analyzed the power of various “things” in IoT products in enhancing the purchase intention
by improving the functional and emotional experience. Stary [256] stressed that IoB would
transform the business and organization space with its choreographic intelligence. IoB is
now in its infancy, and its success coincides with large-scale IoT deployments with a high
level of user acceptance.

5.2. Pandemic Management

Today’s world is witnessing the devastation of the COVID-19 pandemic. The WHO
categorically said several times that the world response could be far better than what we
did and are doing. Whether in developed countries like Italy and the US or developing
countries like India and Brazil, most health-care systems were underprepared and already
overburdened. From the prism of sophisticated technologies, IoT can be used effectively to
monitor and control the COVID-19 pandemic. IoT infrastructure coupled with intelligence
can be used to address challenges during the lockdowns, social distancing, contact tracing,
health-care monitoring, prescreening, remote meeting, anytime and anywhere accessibility,
etc. [257]. IoT can play a significant role in providing virtual health-care (contactless)
tools and telemedicine to the masses, which will eventually help in achieving the goals of
Healthcare 4.0 [258]. For example, Smart Field Hospital in Wuhan used IoT and AI-based
applications to help health-care workers relax. Robots and IoT devices helped to perform
contactless body temperature monitoring, cleaning, disinfecting, etc. [259].
On the other hand, IoT sensors can help to track infection by forming the web of human
nodes and their connections. However, it has some serious privacy concerns, which need to
be addressed [260]. Why has the world messed up in managing and controlling COVID-19?
The answer is because human decision-making is slow and biased. To address this issue,
Alam et al. [261] proposed iResponse, an intelligent IoT-enabled system for autonomous
COVID-19 pandemic management. The authors demonstrated through iResponse that the
fusion of IoT and intelligence could help break the chain of infection, cure development,
treatment, resource planning, pandemic analytics, and decision-making. Still, we need to
deploy IoT infrastructure on a large scale around the world to exploit its benefits.

5.3. Connected Autonomous Vehicles

IoT will connect AVs and will help in developing the driving cognitive. However,
connected AV development is in its infancy, and a lot depends on how the adopter’s
willingness to accept the change and pricing of these vehicles, as examined by Talebian
and Mishra [262], as well as such issues as pedestrian detection, intersection navigation,
communications, collision avoidance, and security. One of the first of this kind of work was
done by Alam et al., who developed TAAWUN [263]. It uses connected vehicle data and
prediction to enhance its driving-scene understanding. The core concept used in TAAWUN
can also use IoT infrastructure in the future and sense data for prediction. AVs have recently
suffered deadly crashes during the testing phase [264,265]. This shows that ML algorithms
used by AVs are not yet matured for real-world challenges. After TAAWUN, there are now
a few words examining the benefits of connected AVs. Elliott et al. [266] critically discussed
recent advances in connected AVs, focusing on five major areas: intersection navigation,
pedestrian detection, collision avoidance, communications, and security. Safety is one of
the foremost goals of any autonomous technology. Addressing the safety aspect, Ye and
Electronics 2022, 11, 2676 20 of 33

Yamamoto [267] critically analyzed the impact of connected AVs in providing a hassle-free
and smooth driving experience with enhanced safety.

5.4. Edge and Fog Computing

Edge and fog computing push data intelligence and its processing closer to the nodes
from where data is sensed or required, as depicted in Figure 9. Edge computing brings
computational power closer to the sensing data than sending it to a remote cloud [268].
This results in efficient speed and enhanced data-transport performance of devices and
applications. Fog computing, an emergent architecture, can be termed a subset of the edge
computing paradigm [269]. Fog computing enables the cloud to be closer to the smart
objects that generate data and actuators that act on sensed data. It defines the standards
related to edge computing, data transfer, storage, computation, and networking [270]. Edge
and fog computing are enabling technologies that will help IoT infrastructure to assist
smart applications and serve the bigger objective of smart cities around the globe.

Figure 9. Edge and fog computing landscape.

5.5. Lightweight Deep Learning

Deep learning is a representational learning model that mimics the neural system of
humans. It takes raw data as input and automatically discovers representations required
to make predictions. Deep learning tries to model higher-level data abstractions. A deep-
learning model can have several layers between input and output, which helps it to think.
An intriguing fact about deep learning is that layers of features are learned from data auto-
matically. LeCun, the director of AI research at Facebook [271]., stated that deep learning
would see many near-future successes because of two critical factors: (1) limited engineer-
ing by hand required and (2) taking leverage from enhanced computational resources
and data availability. However, deep learning has a few issues, such as these algorithms
consuming too many resources like processing power and energy resources. From an
IoT perspective, we need to exploit the power of deep learning at several levels, such as
(1) cloud, (2) fog, and (3) edge [39,272]. Plenty of processing resources are available at the
cloud level, which the deep-learning algorithms can consume. Further going downwards
to fog, the availability of processing resources decreases significantly, whereas edge has
Electronics 2022, 11, 2676 21 of 33

minimum processing resources, so unsuitable for conventional deep-learning algorithms.

In addition, these deep networks must be capable of perceiving the environment from fewer
data. Recent deep-learning trends show fundamental understanding among researchers
about the need for lightweight deep-learning algorithms for IoT [273]. Several advances
reflect this trend, such as Alibaba introducing the open-source mobile neural network [274],
a lightweight deep-learning model for HAR using smart “things” on edge [275], Mobi-
Face, ShuffleFaceNet for face recognition on mobile devices [274,276], lightweight machine
learning for IoT Systems (LIMITS) [277], CardioXNet, a lightweight deep-learning frame-
work for heart-disease prediction [278], lightweight deep learning-based virtual vision
sensing technology [279], and lightweight convolutional neural networks (CNNs) [280,281].
In the future, we expect more development in this direction, as all or most IoT devices
cannot match conventional deep-learning systems in terms of processing, memory, or
power requirements.

6. Challenges
The IoT paradigm is a perfect candidate to bridge the gap between the real and digital
worlds by developing a hyperconnected world. However, it needs to overcome and manage
massive technological and nontechnological challenges. Based on several literature reviews
on IoT [282–285], we identified four classes of challenges: (1) technological, (2) individual,
(3) business, and (4) society. In the proceeding subsections, we discuss them in brief.

6.1. Technological Challenges

The extensive deployment of IoT systems is still a distant reality. However, they are
being implemented in bits and pieces around the world. Sensing, connectivity, actuators,
and security are four significant technologies fused to make IoT. The world still faces
connectivity issues. Mobile connectivity and internet availability are obstacles, particularly
in low-income countries. Another problem is that cross-platform capability is not much
there in the present IoT platform, resulting in its slow acceptance [286,287], Nowadays, the
IoT landscape is primarily based on a client–server architecture, which is not a feasible
option for the future because of increased latency, maximized energy consumption, single-
point failure, and security vulnerabilities. To tackle this, edge and fog computing platforms
came into being, albeit most of them are in their infancy and facing challenges like network
bandwidth, latency, accessibility, control, and management [288]. A bigger challenge
posing a significant threat to the large-scale acceptance of IoT systems is the security of
these “things” [285]. IoT is about decentralized edge devices where devices will connect
with several strange things that are more prone to cyberattacks. Universal standards
for IoT devices are not there for authentication and authorization, which adds further to
security vulnerabilities. Security is one of the most critical issues IoT needs to address for a
successful and acceptable paradigm.

6.2. Individual Challenges

Different people will use IoT for a diverse set of needs. The aim is to improve daily lives
through sophisticated automation and an intelligent environment. Due to IoT’s “anytime
and anywhere” property, the foremost concern is privacy as it varies from individual to
individual, resulting in more challenging scenarios that devices and servers need to handle.
Sen et al. [289], critically discussed the ways to preserve privacy and emerging related
trends in IoT. Another hurdle to IoT’s success is the acceptability of IoT among the public.
IoT will change how we look at our daily life. The point of interest is how compatible our
cognitive needs are with these changes [290]. Beştepe and Yildirim [291] analyzed how
public acceptability is essential for smart cities, which at their core use IoT infrastructure to
achieve sustainability. To increase IoT’s acceptability, we need to educate and train people
to make them aware of the benefits these sophisticated applications and services bring and
how they can revolutionize our daily lives towards more prosperity.
Electronics 2022, 11, 2676 22 of 33

6.3. Business Challenges

IoT offers enormous business opportunities in manufacturing, applications, and ser-
vices. However, it cannot compete with the hype that was created. Businesses are facing
challenges including but not limited to lack of universal platforms, lack of industry stan-
dards, compatibility connectivity, data collection, and security issues [292,293]. A skill
shortage of IoT experts is also one major issue. Other than this, global market anoma-
lies like COVID-19 [294], a global computer chip shortage [295], and the Russia–Ukraine
war [296] have slowed the pace and reduced the interest of businesses and governments in
IoT developments as priorities shifted.

6.4. Society Challenges

As a society, we need a prosperous and sustainable living environment. IoT can help
significantly by providing actionable decision-making support [297]. However, the critical
question is: Are we equipped today as a society to use IoT? And are we ready to accept
the cognitive changes it will bring to our daily lives? These essential questions will shape
public acceptance of IoT applications and services [298]. It will increase the demand side,
encouraging industries to put serious efforts into the large-scale deployment of IoT, its
applications, and services. Further, the digital divide is another major challenge today our
societies suffer due to the imbalance in global economic growth. If we want to exploit IoT
to its full potential, we must address the issues discussed above.

7. Conclusions
IoT is far more mature now. More IoT applications are in practical use. Individuals,
governments, and businesses have shown a keen interest in leveraging IoT’s opportunities.
An important question remains: How will IoT learn and think to provide a high degree of
automation? The answer comes from other branches of computer science that understand
and act like humans with the help of ML. In this paper, rather than doing a classical review
of literature, we tried highlighting the importance of ML for IoT’s success and diverse
ML-powered IoT applications. We classify ML developments in IoT from three perspectives:
data, application, and industry. The literature reviewed is wholly or partially applicable to
the IoT ecosystem. Further, we identified and discussed emerging IoT trends, including
Internet of Behavior (IoB), pandemic management, edge and fog computing, connected
autonomous vehicles, and lightweight deep learning, with a primary focus on machine
learning to develop futuristic and sustainable solutions. Despite IoT’s ability to transform
our present-day societies into smarter and more sustainable ones, it has to overcome a set
of challenges, e.g., technological, individual, business, and those related to our societies.
We conclude that ML developments in IoT will revolve around currently available
and well-established ML methods, at least in the short term. However, in the future, we
can see a fully autonomous IoT ecosystem with embedded intelligence capabilities that will
be a tricky development from an ML point of view regarding device data and processing
abilities. With the help of this work, the reader can see what ML means to IoT, how ML is
used with IoT, and what the prospects of ML in IoT can be.

Author Contributions: Conceptualization, J.B. and F.A.; literature classification, J.B., A.D., S.M.A.
and I.K.N.; software, J.B., F.A. and S.M.A.; validation, F.A.; literature analysis, J.B. and A.D.; resources,
R.M., I.K.N. and M.B.; writing—original draft preparation, J.B., F.A. and R.M.; writing—review and
editing, J.B., F.A. and A.D.; visualization, J.B., A.D. and S.M.A.; supervision, F.A., R.M. and M.B;
project administration, F.A., R.M. and M.B. All authors have read and agreed to the published version
of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: All the data used in this work are publicly available and appropriately cited.
Conflicts of Interest: The authors declare no conflict of interest.
Electronics 2022, 11, 2676 23 of 33

References
1. Atzori, L.; Iera, A.; Morabito, G. The Internet of Things: A Survey. Comput. Netw. 2010, 54, 2787–2805. [CrossRef]
2. Statista IoT Connected Devices Worldwide 2019–2030. Available online: https://www.statista.com/statistics/1183457/iot-
connected-devices-worldwide/ (accessed on 25 June 2022).
3. Transforma Insights Global IoT Market to Grow to 24.1 Billion Devices in 2030, Generating $1.5 Trillion Annual Revenue. Available
online: https://transformainsights.com/news/iot-market-24-billion-usd15-trillion-revenue-2030 (accessed on 25 June 2022).
4. Kar, I. The Top-Country Early-Adopters of the Internet of Things, Ranked Google Wants to Have Drones Buzzing around
Offices. Available online: http://qz.com/648863/the-top-country-early-adopters-of-the-internet-of-things-ranked/ (accessed
on 2 June 2022).
5. Mattern, F.; Floerkemeier, C. From the Internet of Computers to the Internet of Things. In From Active Data Management to
Event-Based Systems and More; Springer: Berlin/Heidelberg, Germany, 2010; Volume 6462, pp. 242–259. [CrossRef]
6. Tanwar, S.; Garg, J.; Gupta, M.; Rana, A. Opportunities and Challenges in Machine Learning with IoT. Available online:
https://wandb.ai/site?utm_source=bing&utm_medium=cpc&utm_campaign=Conversions%3A+ML+Tools&utm_content=
Ad+group+1&msclkid=4b1aa96029de11b0fa09ca68ba118cd4&utm_term=machine%20learning%20tools (accessed on 25
June 2022).
7. Guo, B.; Zhang, D.; Yu, Z.; Liang, Y.; Wang, Z.; Zhou, X. From the Internet of Things to Embedded Intelligence. World Wide Web
2013, 16, 399–420. [CrossRef]
8. Guo, B.; Zhang, D.; Wang, Z. Living with Internet of Things: The Emergence of Embedded Intelligence. In Proceedings of the
2011 International Conference on Internet of Things and 4th International Conference on Cyber, Physical and Social Computing,
Dalian, China, 19–22 October 2011; pp. 297–304. [CrossRef]
9. Jain, R. Recent Machine Learning Applications to Internet of Things (IoT). Abstract: Table of Contents: 1–19. Available online:
https://www.cs.wustl.edu/~{}jain/cse570-15/ftp/iot_ml/#sec7 (accessed on 2 June 2022).
10. Keim, D.A.; Munzner, T.; Rossi, F.; Verleysen, M. Bridging Information Visualization with Machine Learning (Dagstuhl Seminar
15101). Dagstuhl. Rep. 2015, 5, 7. [CrossRef]
11. Davarzani, B.L.; Purdy, M. The Internet of Things Is Now a Thing 2015. Available online: https://ssir.org/articles/entry/the_
internet_of_things_is_now_a_thing (accessed on 2 June 2022).
12. Purdy, M.; Ladan, D. The Growth Game-Changer: How the Industrial Internet of Things Can Drive Progress and Prosperity;
2015. Available online: https://www.bl.uk/business-and-management/collection-items/suppressed-by-publisher/accenture/
growth-gamechanger-how-the-industrial-internet-of-things-can-drive-progress-and-prosperity (accessed on 2 June 2022).
13. Kearney, A.T. The Internet of Things: A New Path to European Prosperity; 2016. Available online: https://www.fr.kearney.com/
digital-transformation/article?/a/the-internet-of-things-a-new-path-to-european-prosperity (accessed on 2 June 2022).
14. Zaslavsky, A.; Perera, C.; Georgakopoulos, D. Sensing as a Service and Big Data. In Proceedings of the International Conference
on Advances in Cloud Computing (ACC), Bangalore, India, 4–6 July 2012; pp. 21–29.
15. Isik, O.; Jones, M.; Sidorova, A. ‘Big data’, the ‘internet of things’ and the ‘internet of signs’ daniel. Intell. Syst. Account. Financ.
Manag. 2011, 176, 161–176. [CrossRef]
16. Sun, Y.; Song, H.; Jara, A.; Bie, R. Internet of Things and Big Data Analytics for Smart and Connected Communities. IEEE Access
2016, 4, 766–773. [CrossRef]
17. Lindstrom, M. Small Data: The Tiny Clues That Uncover Huge Trends; St. Martin’s Press: London, UK, 2016; ISBN 9781250080684.
18. Kavis, M. Forget Big Data—Small Data Is Driving. Forbes 2016, 1–5. Available online: http://www.forbes.com/sites/mikekavis/
2015/02/25/forget-big-data-small-data-is-driving-the-internet-of-things/#3de078be661b. (accessed on 2 June 2022).
19. Chen, F.; Deng, P.; Wan, J.; Zhang, D.; Vasilakos, A.V.; Rong, X. Data Mining for the Internet of Things: Literature Review and
Challenges. Int. J. Distrib. Sens. Netw. 2015, 11, 431047. [CrossRef]
20. Qin, Y.; Sheng, Q.Z.; Falkner, N.J.G.; Dustdar, S.; Wang, H.; Vasilakos, A.V. When Things Matter: A Survey on Data-Centric
Internet of Things. J. Netw. Comput. Appl. 2016, 64, 137–153. [CrossRef]
21. Abraham, A.; Muda, A.K.; Choo, Y.H. Pattern Analysis, Intelligent Security and the Internet of Things; Springer: Cham, Switzerland,
2015; Volume 355. [CrossRef]
22. Zhang, L.; Jeong, D.; Lee, S. Data Quality Management in the Internet of Things. Sensors 2021, 21, 5834. [CrossRef]
23. Aggarwal, C.C.; Ashish, N.; Sheth, A. The Internet of Things: A Survey from the Data-Centric Perspective. In Managing and
Mining Sensor Data; Springer: Boston, MA, USA, 2013; pp. 383–428. [CrossRef]
24. Abu-Elkheir, M.; Hayajneh, M.; Ali, N.A. Data Management for the Internet of Things: Design Primitives and Solution. Sensors
2013, 13, 15582–15612. [CrossRef]
25. Torrance, M. Better Algorithms Beat More Data—And Here’ s Why. Available online: http://allthingsd.com/20121128/better-
algorithms-beat-more-data-and-heres-why/ (accessed on 4 July 2022).
26. Cleland, S. Google’s " Infringenovation " Secrets. Available online: http://www.forbes.com/sites/scottcleland/2011/10/03
/googles-infringenovation-secrets/#3e263e2373de (accessed on 6 July 2022).
27. Domingos, P. A Few Useful Things to Know about Machine Learning. Commun. ACM 2012, 55, 78. [CrossRef]
28. Halevy, A.; Norvig, P.; Pereira, F. The Unreasonable Effectiveness of Data. IEEE Intell. Syst. 2009, 24, 8–12. [CrossRef]
Electronics 2022, 11, 2676 24 of 33

29. Sessions, V.; Valtorta, M. The Effects of Data Quality on Machine Learning Algorithms. ICIQ. Available online: http:
//mitiq.mit.edu/ICIQ/Documents/IQ%20Conference%202006/papers/The%20Effects%20of%20Data%20Quality%20on%20
Machine%20Learning%20Algorithms.pdf (accessed on 2 June 2022).
30. Cortes, C.; Jackel, L.; Chiang, W. Limits on Learning Machine Accuracy Imposed by Data Quality. Available online: https:
//www.aaai.org/Papers/KDD/1995/KDD95-007.pdf (accessed on 2 June 2022).
31. Ambasna-Jones, M. Will Automation and the Internet of Things Lead to Mass Unemployment? Available online: https://www.
theguardian.com/media-network/2015/may/27/internet-of-things-automation-unemployment (accessed on 20 May 2022).
32. Stewart, H. Robot Revolution: Rise of “thinking” Machines Could Exacerbate Inequality. Guardian 2015. Available online:
https://cacm.acm.org/news/194049-robot-revolution-rise-of-thinking-machines-could-exacerbate-inequality/fulltext (accessed
on 20 May 2022).
33. Grigorik, I. Intuition & Data-Driven Machine Learning. Available online: https://www.igvita.com/2011/04/20/intuition-data-
driven-machine-learning/ (accessed on 10 May 2022).
34. Banko, M.; Brill, E. Mitigating the Paucity-of-Data Problem: Exploring the Effect of Training Corpus Size on Classifier Performance
for Natural Language Processing. Comput. Linguist. 2001, 2011, 2–6. [CrossRef]
35. Moreno-Cano, V.; Terroso-Saenz, F.; Skarmeta-Gomez, A.F. Big Data for IoT Services in Smart Cities. In Proceedings of the 2015
IEEE 2nd World Forum on Internet of Things (WF-IoT), Milan, Italy, 14–16 December 2015; pp. 418–423.
36. Shanthamallu, U.S.; Spanias, A.; Tepedelenlioglu, C.; Stanley, M. A Brief Survey of Machine Learning Methods and Their Sensor
and IoT Applications. In Proceedings of the 2017 8th International Conference on Information, Intelligence, Systems Applications
(IISA), Larnaca, Cyprus, 27–30 August 2017; pp. 1–8.
37. Sharma, K.; Nandal, R. A Literature Study on Machine Learning Fusion with IoT. In Proceedings of the International Conference
on Trends in Electronics and Informatics, ICOEI, Tirunelveli, India, 23–25 April 2019; pp. 1440–1445. [CrossRef]
38. Siow, E.; Tiropanis, T.; Hall, W. Analytics for the Internet of Things: A Survey. ACM Comput. Surv. 2018, 51, 1–36. [CrossRef]
39. Mohammadi, M.; Al-Fuqaha, A.; Sorour, S.; Guizani, M. Deep Learning for IoT Big Data and Streaming Analytics: A Survey.
IEEE Commun. Surv. Tutor. 2018, 20, 2923–2960. [CrossRef]
40. Mahdavinejad, M.S.; Rezvan, M.; Barekatain, M.; Adibi, P.; Barnaghi, P.; Sheth, A.P. Machine Learning for Internet of Things Data
Analysis: A Survey. Digit. Commun. Netw. 2018, 4, 161–175. [CrossRef]
41. Alam, F.; Mehmood, R.; Member, S.; Katib, I.; Nasser, N.; Albogami, N.N.N.; Albeshri, A. Data Fusion and IoT for Smart
Ubiquitous Environments: A Survey. IEEE Access 2017, 5, 9533–9554. [CrossRef]
42. De Morais, C.M.; Sadok, D.; Kelner, J. An IoT Sensor and Scenario Survey for Data Researchers. J. Braz. Comput. Soc. 2019, 25, 4.
[CrossRef]
43. Wang, R.; Liu, L.; Shi, W. HydraSpace: Computational Data Storage for Autonomous Vehicles. In Proceedings of the 2020 IEEE
6th International Conference on Collaboration and Internet Computing, CIC, Atlanta, GA, USA, 1–3 December 2020; pp. 70–77.
[CrossRef]
44. Wan, C. Storage and the IoT: What Kind of Storage Is Needed, and Where? Available online: https://blog.westerndigital.com/
storage-iot-kind-storage-needed/ (accessed on 2 June 2022).
45. Llanor Alleyne Best IoT Device Management Platforms & Software. 2022. Available online: https://www.itbusinessedge.com/
cloud/iot-device-management-platorms/ (accessed on 2 July 2022).
46. Scully, P.; Lueth, K.L. Guide to IoT Solution Development. Available online: https://iot-analytics.com/wp/wp-content/uploads/
2016/09/White-paper-Guide-to-IoT-Solution-Development-September-2016-vf.pdf (accessed on 1 June 2022).
47. Hodge, V.J.; Austin, J. A Survey of Outlier Detection Methodoligies. Artif. Intell. Rev. 2004, 22, 85–126. [CrossRef]
48. Van Den Broeck, J.; Cunningham, S.A.; Eeckels, R.; Herbst, K. Data Cleaning: Detecting, Diagnosing, and Editing Data
Abnormalities. PLoS Med. 2005, 2, 0966–0970. [CrossRef]
49. Ben-gal, I. Outlier Detection. In Data Mining and Knowledge Discovery Handbook; Springer: Boston, MA, USA, 2005; pp. 131–146.
[CrossRef]
50. Kempf, J.; Arkko, J.; Beheshti, N.; Yedavalli, K. Thoughts on Reliability in the Internet of Things. In Proceedings of the
Interconnecting Smart Objects with the Internet Workshop, Prague, Czech Republic, 25 March 2011; pp. 1–4.
51. Mahmood, M.A.; Seah, W.K.G.; Welch, I. Reliability in Wireless Sensor Networks: A Survey and Challenges Ahead. Comput.
Netw. 2015, 79, 166–187. [CrossRef]
52. Andreas, W.; Karl, H. Data Transport Reliability in Wireless Sensor Networks; St. Petersburg State University: St. Petersburg, Russia,
2005; p. 28.
53. Qiu, J.; Wu, Q.; Ding, G.; Xu, Y.; Feng, S. A Survey of Machine Learning for Big Data Processing. EURASIP J. Adv. Signal Process.
2016, 27, 327–336. [CrossRef]
54. Fawzy, A.; Mokhtar, H.M.O.; Hegazy, O. Outliers Detection and Classification in Wireless Sensor Networks. Egypt. Inform. J. 2013,
14, 157–164. [CrossRef]
55. Shahid, N.; Naqvi, I.H.; Qaisar, S. Bin Characteristics and Classification of Outlier Detection Techniques for Wireless Sensor
Networks in Harsh Environments: A Survey. Artif. Intell. Rev. 2012, 43, 193–228. [CrossRef]
56. Meratnia, N.; Havinga, P. Outlier Detection Techniques for Wireless Sensor Networks: A Survey. IEEE Commun. Surv. Tutor. 2010,
12, 159–170. [CrossRef]
Electronics 2022, 11, 2676 25 of 33

57. Garcia-font, V.; Garrigues, C. A Comparative Study of Anomaly Detection Techniques for Smart City Wireless Sensor Networks.
Sensors 2016, 16, 868. [CrossRef]
58. Alsheikh, M.A.; Lin, S.W.; Niyato, D.; Tan, H. Machine Learning in Wireless Sensor Networks: Algorithms, Strategies, and
Applications. IEEE Commun. Surv. Tutor. 2014, 16, 1996–2018. [CrossRef]
59. Alghanmi, N.; Alotaibi, R.; Buhari, S.M. Machine Learning Approaches for Anomaly Detection in IoT: An Overview and Future
Research Directions. Wirel. Pers. Commun. 2021, 122, 2309–2324. [CrossRef]
60. Cook, A.A.; Misirli, G.; Fan, Z. Anomaly Detection for IoT Time-Series Data: A Survey. IEEE Internet Things J. 2020, 7, 6481–6494.
[CrossRef]
61. Diro, A.; Chilamkurti, N.; Nguyen, V.D.; Heyne, W. A Comprehensive Study of Anomaly Detection Schemes in IoT Networks
Using Machine Learning Algorithms. Sensors 2021, 21, 8320. [CrossRef]
62. Al Samara, M.; Bennis, I.; Abouaissa, A.; Lorenz, P. A Survey of Outlier Detection Techniques in IoT: Review and Classification.
J. Sens. Actuator Netw. 2022, 11, 4. [CrossRef]
63. Jordaan, E.E.M.; Smits, G.G.F. Robust Outlier Detection Using SVM Regression. In Proceedings of the 2004 IEEE International
Joint Conference on Neural Networks (IEEE Cat. No.04CH37541), Budapest, Hungary, 25–29 July 2004; Volume 3, pp. 2017–2022.
[CrossRef]
64. Yang, Z.Y.Z.; Meratnia, N.; Havinga, P. An Online Outlier Detection Technique for Wireless Sensor Networks Using Unsupervised
Quarter-Sphere Support Vector Machine. In Proceedings of the 2008 International Conference on Intelligent Sensors, Sensor
Networks and Information Processing, Sydney, Australia, 15–18 December 2008; pp. 151–156. [CrossRef]
65. Rajasegarar, S.; Leckie, C.; Palaniswami, M.; Bezdek, J.C. Quarter Sphere Based Distributed Anomaly Detection in Wireless Sensor
Networks. In Proceedings of the ICC ’07: 2007 IEEE International Conference on Communications, Glasgow, UK, 24–28 June
2007; Volume 1, pp. 3864–3869. [CrossRef]
66. Münz, G.; Li, S.; Carle, G. Traffic Anomaly Detection Using K-Means Clustering. Available online: https://www.semanticscholar.
org/paper/Traffic-Anomaly-Detection-Using-K-Means-Clustering-M%C3%BCnz-Li/634e2f1a20755e7ab18e8e8094f48e140a3
2dacd (accessed on 2 June 2022).
67. Ahmed, M.; Seraj, R.; Islam, S.M.S. The K-Means Algorithm: A Comprehensive Survey and Performance Evaluation. Electronics
2020, 9, 1295. [CrossRef]
68. Aubaidan, B.; Mohd, M.; Albared, M.; Author, F. Comparative Study of K-Means and K-Means++ Clustering Algorithm on Crime
Domain. J. Comput. Sci. 2014, 10, 1197–1206. [CrossRef]
69. Statman, A.; Rozenberg, L.; Feldman, D. K-Means+++: Outliers-Resistant Clustering. Algorithms 2020, 13, 311. [CrossRef]
70. Titouna, C.; Aliouat, M.; Gueroui, M. Outlier Detection Approach Using Bayes Classifiers in Wireless Sensor Networks. Wirel.
Pers. Commun. 2015, 85, 1009–1023. [CrossRef]
71. Ting, J.A.; D’Souza, A.; Schaal, S. Automatic Outlier Detection: A Bayesian Approach. In Proceedings of the 2007 IEEE
International Conference on Robotics and Automation, Rome, Italy, 10–14 April 2007; pp. 2489–2494. [CrossRef]
72. Parto, M.; Saldana, C.; Kurfess, T. Real-Time Outlier Detection and Bayesian Classification Using Incremental Computations for
Efficient and Scalable Stream Analytics for IoT for Manufacturing. Procedia Manuf. 2020, 48, 968–979. [CrossRef]
73. Lam, P.; Wang, L.; Ngan, H.Y.T.; Yung, N.H.C.; Yeh, A.G.O.; Wang, L.; Ngan, H.Y.T.; Yung, N.H.C.; Yeh, A.G.O. Outlier Detection
in Large-Scale Traffic Data by Naïve Bayes Method and Gaussian Mixture Model Method. Electron. Imaging 2017, 29, 73–78.
[CrossRef]
74. Alam, F.; Mehmood, R.; Katib, I.; Albeshri, A. Analysis of Eight Data Mining Algorithms for Smarter Internet of Things (IoT). In
Proceedings of the International Workshop on Data Mining in IoT Systems (DaMIS 2016), London, UK, 19–22 September 2016;
pp. 437–442. [CrossRef]
75. Luo, T.; Nagarajan, S.G. Distributed Anomaly Detection Using Autoencoder Neural Networks in WSN for IoT. In Proceedings of
the 2018 IEEE International Conference on Communications (ICC), Kansas City, MO, USA, 20–24 May 2018; pp. 1–6.
76. Diro, A.A.; Chilamkurti, N. Distributed Attack Detection Scheme Using Deep Learning Approach for Internet of Things. Future
Gener. Comput. Syst. 2018, 82, 761–768. [CrossRef]
77. Zhao, K.; Pan, K.; Zhang, B. Cooperative Transmission in Mobile Wireless Sensor Networks with Multiple Carrier Frequency
Offsets: A Double-Differential Approach. Math. Probl. Eng. 2014, 2014, 853821. [CrossRef]
78. Ding, Z.; Mei, G.; Cuomo, S.; Li, Y.; Xu, N. Comparison of Estimating Missing Values in IoT Time Series Data Using Different
Interpolation Algorithms. Int. J. Parallel Program. 2018, 48, 534–548. [CrossRef]
79. Farhangfar, A.; Kurgan, L.; Dy, J. Impact of Imputation of Missing Values on Classification Error for Discrete Data. Pattern
Recognit. 2008, 41, 3692–3705. [CrossRef]
80. Wu, X.; Kumar, V.; Quinlan, J.R.; Ghosh, J.; Yang, Q.; Motoda, H.; Mclachlan, G.J.; Ng, A.; Liu, B.; Yu, P.S.; et al. Top 10 Algorithms
in Data Mining. Knowl. Inf. Syst. 2008, 14, 1–37. [CrossRef]
81. Wong, L.Z.; Chen, H.; Lin, S.; Chen, D.C. Imputing Missing Values in Sensor Networks Using Sparse Data Representations.
In Proceedings of the 17th ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile
Systems—MSWiM ‘14, Montreal, QC, Canada, 21–26 September 2014; pp. 227–230. [CrossRef]
82. Ni, J.; Li, L.; Qiao, F.; Wu, Q. A GS-MPSO-WKNN Method for Missing Data Imputation in Wireless Sensor Networks Monitoring
Manufacturing Conditions. Trans. Inst. Meas. Control 2014, 36, 1083–1092. [CrossRef]
Electronics 2022, 11, 2676 26 of 33

83. Pan, R.; Yang, T.; Cao, J.; Lu, K.; Zhang, Z. Missing Data Imputation by K Nearest Neighbours Based on Grey Relational Structure
and Mutual Information. Appl. Intell. 2015, 43, 614–632. [CrossRef]
84. Zhang, S. Nearest Neighbor Selection for Iteratively KNN Imputation. J. Syst. Softw. 2012, 85, 2541–2552. [CrossRef]
85. Weston, J. Support Vector Machine (and Statistical Learning Theory) Tutorial. Bioinformatics 2005, 21, 631–643. [CrossRef]
86. Pelckmans, K.; De Brabanter, J.; Suykens, J.A.K.; De Moor, B. Handling Missing Values in Support Vector Machine Classifiers.
Neural Netw. 2005, 18, 684–692. [CrossRef]
87. Yang, B.; Janssens, D.; Ruan, D.; Cools, M.; Bellemans, T.; Wets, G. A Data Imputation Method with Support Vector Machines for
ActivityBased Transportation Models. Adv. Intell. Soft Comput. 2012, 122, 249–257. [CrossRef]
88. Yan, X.Z.; Xie, H.; Wan, T.; Luo, Q.H. A Data Predicting Method Based on LS-SVR in Wireless Sensor Network. In Proceedings
of the 2010 International Conference on Broadcast Technology and Multimedia Communication, Hong Kong, China, 28–30
December 2010; p. 1.
89. Zou, C. Node Data Prediction Based on SVR in Wireless Sensor Network. J. Comput. Appl. 2010, 30, 127–129. [CrossRef]
90. Breiman, L.E.O. Random Forests. Mach. Learn. 2001, 45, 5–32. [CrossRef]
91. Fortran Breiman and Cutler’s Random Forests for Classification and Regression. 2015. Available online: http://stat-www.
berkeley.edu/users/breiman/RandomForests (accessed on 2 June 2022).
92. Ishioka, T. Imputation of Missing Values for Semi-Supervised Data Using the Proximity in Random Forests. In Proceedings of
the 14th International Conference on Information Integration and Web-Based Applications & Services, Munich, Germany, 2–4
December 2012; pp. 319–322. [CrossRef]
93. Grzymala-busse, J.W.; Hu, M.; York, N. A Comparison of Several Approaches to Missing Attribute Values in Data Mining. In
International Conference on Rough Sets and Current Trends in Computing; Springer: Berlin/Heidelberg, Germany, 2000; pp. 378–385.
94. Batista, G.E.A.P.A.; Monard, M.C. A Study of K-Nearest Neighbour as an Imputation Method. In Proceedings of the Soft
Computing Systems—Design, Management and Applications, HIS Conference, Santiago, Chile, 1–4 December 2002.
95. Song, Q.; Shepperd, M.; Chen, X.; Liu, J. Can k -NN Imputation Improve the Performance of C4.5 with Small Software Project
Data Sets? A Comparative Evaluation. J. Syst. Softw. 2008, 81, 2361–2370. [CrossRef]
96. Mehala, B.; Vivekanandan, K.; Thangaiah, P.R.J. An Analysis on K-Means Algorithm as an Imputation Method to Deal with
Missing Values. Asian J. Inf. Technol. 2008, 7, 434–441.
97. Schmitt, P.; Mandel, J.; Guedj, M. A Comparison of Six Methods for Missing Data Imputation. Biom. Biostat. 2015, 6, 1. [CrossRef]
98. Li, D.; Deogun, J.; Spaulding, W.; Shuart, B. Towards Missing Data Imputation: A Study of Fuzzy K-Means Clustering Method.
Rough Sets Curr. Trends Comput. 2004, 3066, 573–579. [CrossRef]
99. Hruschka, E.R.; Hruschka, E.R., Jr.; Ebecken, N.F.F. Towards Efficient Imputation by Nearest-Neighbors: A Clustering-Based
Approach. In Australasian Joint Conference on Artificial Intelligence; Springer: Berlin/Heidelberg, Germany, 2004; pp. 513–525.
100. Patil, B.M.; Joshi, R.C.; Toshniwal, D. Missing Value Imputation Based on K-Mean Clustering with Weighted Distance. Commun.
Comput. Inf. Sci. 2010, 94, 600–609. [CrossRef]
101. Chi, J.T.; Chi, E.C.; Baraniuk, R.G. A Method for K-Means Clustering of Missing Data; Rice University: Houston, TX, USA, 2016.
102. Aydilek, I.B.; Arslan, A. A Hybrid Method for Imputation of Missing Values Using Optimized Fuzzy C-Means with Support
Vector Regression and a Genetic Algorithm. Inf. Sci. 2013, 233, 25–35. [CrossRef]
103. Kian, L.; Chu, L.; Way, L. Autonomous and Deterministic Supervised Fuzzy Clustering with Data Imputation Capabilities. Appl.
Soft Comput. 2011, 11, 1117–1125. [CrossRef]
104. Bu, F.; Chen, Z.; Zhang, Q.; Yang, L.T. Incomplete High-Dimensional Data Imputation Algorithm Using Feature Selection and
Clustering Analysis on Cloud. J. Supercomput. 2012, 72, 2977–2990. [CrossRef]
105. Tian, J.; Yu, B.; Yu, D.; Ma, S. Missing Data Analyses: A Hybrid Multiple Imputation Algorithm Using Gray System Theory and
Entropy Based on Clustering. Appl. Intell. 2014, 40, 376–388. [CrossRef]
106. Rey-del-castillo, P. Fuzzy Min-Max Neural Networks for Categorical Data: Application to Missing Data Imputation. Neural
Comput. Appl. 2012, 21, 1349–1362. [CrossRef]
107. Gautam, C.; Ravi, V. Neurocomputing Data Imputation via Evolutionary Computation, Clustering and a Neural Network.
Neurocomputing 2015, 156, 134–142. [CrossRef]
108. Ennett, C.M.; Frize, M.; Member, S.; Walker, C.R. Imputation of Missing Values by Integrating Neural Networks and Case-Based
Reasoning. In Proceedings of the 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology
Society, Vancouver, BC, Canada, 20–25 August 2008; pp. 4337–4341.
109. Ravi, V.; Krishna, M. A New Online Data Imputation Method Based on General Regression Auto Associative Neural Network.
Neurocomputing 2014, 138, 106–113. [CrossRef]
110. Ultsch, A.; Rolf, S. Imputation of Missing Values by Neural Networks for Data Mining Applications. Knowl. Bases Syst. 2000,
182, 3–6.
111. Saar-Tsechansky, M.; Provost, F. Handling Missing Values When Applying Classification Models. J. Mach. Learn. Res. 2007, 8,
1625–1657.
112. Guyon, I.; Elisseeff, A. An Introduction to Variable and Feature Selection. J. Mach. Learn. Res. 2003, 3, 1157–1182.
113. Bratu, C.V.; Muresan, T.; Potolea, R. Improving Classification Accuracy through Feature Selection. In Proceedings of the 4th
International Conference on Intelligent Computer Communication and Processing, Cluj-Napoca, Romania, 28–30 August 2008;
pp. 25–32. [CrossRef]
Electronics 2022, 11, 2676 27 of 33

114. Chu, C.; Hsu, A.; Chou, K.; Bandettini, P.; Lin, C. Does Feature Selection Improve Classification Accuracy? Impact of Sample Size
and Feature Selection on Classification Using Anatomical Magnetic Resonance Images. Neuroimage 2012, 60, 59–70. [CrossRef]
115. Lin, K.; Lin, C.; Huang, C.; Chang, H.; Yang, C.; Lin, C.; Tang, C.Y.; Hsu, D.F.; Member, S. Feature Selection and Combination
Criteria for Improving Accuracy in Protein Structure Prediction. IEEE Trans. Nanobiosci. 2007, 6, 186–196. [CrossRef]
116. Janecek, A.G.K.; Gansterer, W.N.; Demel, M.A.; Ecker, G.F. On the Relationship between Feature Selection and Classification
Accuracy. In Proceedings of the JMLR Workshop and Conference Proceedings, Antwerp, Belgium, 15 September 2008; pp. 90–105.
117. Fonollosa, J.; Rodríguez-Luján, I.; Trincavelli, M.; Vergara, A.; Huerta, R. Chemical Discrimination in Turbulent Gas Mixtures
with MOX Sensors Validated by Gas Chromatography-Mass Spectrometry. Sensors 2014, 14, 19336–19353. [CrossRef]
118. Ziyatdinov, A.; Fonollosa, J.; Fernández, L.; Gutierrez-Gálvez, A.; Marco, S.; Perera, A. Bioinspired Early Detection through Gas
Flow Modulation in Chemo-Sensory Systems. Sens. Actuators B Chem. 2015, 206, 538–547. [CrossRef]
119. Vergara, A.; Vembu, S.; Ayhan, T.; Ryan, M.A.; Homer, M.L.; Huerta, R. Chemical Gas Sensor Drift Compensation Using Classifier
Ensembles. Sens. Actuators B Chem. 2012, 166–167, 320–329. [CrossRef]
120. Molina, L.C.; Belanche, L.; Nebot, À.; Girona, J. Feature Selection Algorithms: A Survey and Experimental Evaluation. In
Proceedings of the 2002 IEEE International Conference on Data Mining, Maebashi, Japan, 9–12 December 2002.
121. Chizi, B.; Rokach, L.; Maimon, O. A Survey of Feature Selection Techniques. In Encyclopedia of Data Warehousing and Mining,
2nd ed.; Idea Group Publishing: London, UK, 2009; pp. 1888–1895.
122. Amr, T.; Iglesia, B.D. La Survey on Feature Selection. Available online: https://www.arxiv-vanity.com/papers/1510.02892/
(accessed on 2 June 2022).
123. Xiang, J.; Han, X.; Duan, F.; Qiang, Y.; Xiong, X.; Lan, Y.; Chai, H. A Novel Hybrid System for Feature Selection Based on an
Improved Gravitational Search Algorithm and K-NN Method. Appl. Soft Comput. J. 2015, 31, 293–307. [CrossRef]
124. Li, S.; Harner, E.J.; Adjeroh, D.A. Random KNN Feature Selection—A Fast and Stable Alternative to Random Forests. BMC
Bioinform. 2011, 12, 450. [CrossRef] [PubMed]
125. Li, S. Random KNN Classification and Regression. 2015. Available online: https://www.coursehero.com/file/140546671/kNN-
classification-and-regressionpdf/ (accessed on 2 June 2022).
126. Su, M.; Chang, K.; Wei, H.; Lin, C. Feature Weighting and Selection for a Real-Time Network Intrusion Detection System Based on
GA with KNN. Lect. Notes Comput. Sci. 2008, 5075, 195–204. [CrossRef]
127. Chang, Y.; Lin, C. Feature Ranking Using Linear SVM. JMLR Workshop Conf. Proc. 2008, 3, 53–64.
128. Jan, S.U.; Koo, I. A Novel Feature Selection Scheme and a Diversified-Input SVM-Based Classifier for Sensor Fault Classification.
J. Sens. 2018, 2018, 21. [CrossRef]
129. Howcroft, J.; Kofman, J.; Lemaire, E.D. Feature Selection for Elderly Faller Classification Based on Wearable Sensors. J. Neuro Eng.
Rehabil. 2017, 14, 47. [CrossRef]
130. Pandya, R. C5.0 Algorithm to Improved Decision Tree with Feature Selection and Reduced Error Pruning. Int. J. Comput. Appl.
2015, 117, 18–21.
131. Thomas, S.; Bourobou, M.; Yoo, Y. User Activity Recognition in Smart Homes Using Pattern Clustering Applied to Temporal
ANN Algorithm. Sensors 2015, 15, 11953–11971. [CrossRef]
132. Wang, R. AdaBoost for Feature Selection, Classification and Its Relation with SVM, A Review. Phys. Procedia 2012, 25, 800–807.
[CrossRef]
133. Fürst, L.; Fidler, S.; Leonardis, A. Selecting Features for Object Detection Using an AdaBoost-Compatible Evaluation Function.
Pattern Recognit. Lett. 2008, 29, 1603–1612. [CrossRef]
134. Shen, L.; Bai, L. AdaBoost Gabor Feature Selection for Classification. Available online: https://www.semanticscholar.org/
paper/AdaBoost-Gabor-Feature-Selection-for-Classification-Shen-Bai/9a356779ecc242c161cc77caf9052bd91aff1007 (accessed on
2 June 2022).
135. Ma, J.; Cheng, J.C.P. Identifying the Influential Features on the Regional Energy Use Intensity of Residential Buildings Based on
Random Forests. Appl. Energy 2016, 183, 193–201. [CrossRef]
136. Leevy, J.L.; Hancock, J.; Khoshgoftaar, T.M.; Peterson, J.M. IoT Information Theft Prediction Using Ensemble Feature Selection.
J. Big Data 2022, 9, 6. [CrossRef]
137. Hussein, A.Y.; Falcarin, P.; Sadiq, A.T. IoT Intrusion Detection Using Modified Random Forest Based on Double Feature Selection
Methods. Commun. Comput. Inf. Sci. 2022, 1548, 61–78. [CrossRef]
138. Su, J.; He, S.; Wu, Y. Features Selection and Prediction for IoT Attacks. High-Confid. Comput. 2022, 2, 100047. [CrossRef]
139. Taha, A.M.; Mustapha, A.; Chen, S. Naive Bayes-Guided Bat Algorithm for Feature Selection. Sci. World J. 2013, 2013, 325973.
[CrossRef]
140. Feng, G.; Guo, J.; Jing, B.; Sun, T. Feature Subset Selection Using Naive Bayes for Text Classification. Pattern Recognit. Lett. 2015,
65, 109–115. [CrossRef]
141. Ng, A.Y. Feature Selection, L 1 vs. L 2 Regularization, and Rotational Invariance. In Proeedings of the ICML –201904: Proceedings
of the twenty-first international conference on Machine learning, Banff, Alta, Canada, 4–8 July 2004. [CrossRef]
142. Vail, D.L.; Veloso, M.M. Feature Selection for Activity Recognition in Multi-Robot Domains. In Proceedings of the Association for
the Advancement of Artificial Intelligence Conference, Chicago, IL, USA, 13–17 July 2008.
143. Ng, A.Y. Regularization and Feature Selection in Least-Squares Temporal Difference Learning. In Proceedings of the ICML ‘09:
Proceedings of the 26th Annual International Conference on Machine Learning, Montreal, QC, Canada, 14–18 June 2009; p. 94305.
Electronics 2022, 11, 2676 28 of 33

144. Robnik-Sikonja, M.; Kononenko, I. Theoretical and Empirical Analysis of ReliefF and RReliefF. Mach. Learn. 2003, 53, 23–69.
[CrossRef]
145. Zhang, M.; Sawchuk, A.A. A Feature Selection-Based Framework for Human Activity Recognition Using Wearable Multimodal
Sensors. In Proceedings of the ACM Conference on Ubiquitous Computing, Beijing, China, 17–21 September 2011.
146. Gupta, P.; Dallas, T. Feature Selection and Activity Recognition System Using a Single Tri-Axial Accelerometer. IEEE Trans.
Biomed. Eng. Feature 2014, 61, 9294. [CrossRef]
147. Capela, N.A.; Lemaire, E.D.; Baddour, N. Feature Selection for Wearable Smartphone-Based Human Activity Recognition with
Able Bodied, Elderly, and Stroke Patients. PLoS ONE 2015, 10, e0124414. [CrossRef]
148. Zhao, L.; Dong, X. An Industrial Internet of Things Feature Selection Method Based on Potential Entropy Evaluation Criteria.
IEEE Access 2018, 6, 4608–4617. [CrossRef]
149. Whitmore, A.; Agarwal, A.; Da Xu, L. The Internet of Things? A Survey of Topics and Trends. Inf. Syst. Front. 2015, 17, 261–274.
[CrossRef]
150. Xu, L.D.; Member, S.; He, W.; Li, S. Internet of Things in Industries: A Survey. IEEE Trans. Ind. Inform. 2014, 10, 2233–2243.
[CrossRef]
151. United Nations. World’s Population Increasingly Urban with More than Half Living in Urban Areas. Available online: https:
//www.un.org/development/desa/en/news/population/world-urbanization-prospects.html (accessed on 12 June 2022).
152. Kamel Boulos, M.N.; Al-Shorbaji, N.M. On the Internet of Things, Smart Cities and the WHO Healthy Cities. Int. J. Health Geogr.
2014, 13, 10. [CrossRef]
153. High, P. The Top Five Smart Cities in The World. Available online: https://www.forbes.com/sites/peterhigh/2015/03/09/the-
top-five-smart-cities-in-the-world/?sh=461b299667ee (accessed on 25 June 2022).
154. PTI. Smart Cities to Boost Internet of Things Market in 2016: Gartner. Available online: http://timesofindia.indiatimes.com/
tech/tech-news/Smart-cities-to-boost-Internet-of-Things-market-in-2016-Gartner/articleshow/50088586.cms (accessed on 29
June 2022).
155. Ministry of Urban Development. Cities Profile of 20 Smart Cities. Smart Sities Mission. 2016. Available online: https://www.
india.gov.in/smart-cities-mission-portal-ministry-urban-development (accessed on 2 June 2022).
156. Zanella, A.; Bui, N.; Castellani, A.; Vangelista, L.; Zorzi, M. Internet of Things for Smart Cities. Internet Things J. 2014, 1, 22–32.
[CrossRef]
157. Monnier, O. A Smart Grid with the Internet of Things. 2014. Available online: https://www.techonline.com/tech-papers/a-
smarter-grid-with-the-internet-of-things/ (accessed on 2 June 2022).
158. Garg, V.K.; Sharma, S. Role of IoT and Machine Learning in Smart Grid. Lect. Notes Electr. Eng. 2022, 841, 701–715. [CrossRef]
159. Bryant, R.; Hensel, C. From Data to Knowledge to Action: Enabling the Smart Grid. arXiv 2020, arXiv:2008.00055.
160. Zhang, D.; Han, X.; Deng, C. Review on the Research and Practice of Deep Learning and Reinforcement Learning in Smart Grids.
CSEE J. Power Energy Syst. 2018, 4, 362–370. [CrossRef]
161. Wu, L.L.; Kaiser, G.E.; Rudin, C.; Waltz, D.L.; Anderson, R.N.; Boulanger, A.G.; Pooleery, M. Evaluating Machine Learning for
Improving Power Grid Reliability. Comput. Sci. 2011, 2011, 934.
162. Wang, M.; Perera, C.; Jayaraman, P.P.; Zhang, M.; Strazdins, P.; Ranjan, R. City Data Fusion: Sensor Data Fusion in the Internet of
Things. Int. J. Distrib. Syst. Technol. 2015, 7, 15–36. [CrossRef]
163. Rudin, C.; Ertekin, S.; Passonneau, R.; Radeva, A.; Tomar, A.; Xie, B.; Lewis, S.; Riddle, M.; Pangsrivinij, D.; McCormick, T.
Analytics for Power Grid Distribution Reliability in New York City. Interfaces 2014, 44, 364–383. [CrossRef]
164. Aman, S.; Yin, W.; Simmhan, Y.; Prasanna, V. Machine Learning for Demand Forecasting in Smart Grid. Available online:
http://ceng.usc.edu/~{}simmhan/pubs/aman-socalsgs-2011.pdf (accessed on 2 June 2022).
165. Mirowski, P.; Chen, S.; Kam Ho, T.; Yu, C.-N. Demand Forecasting in Smart Grids. Bell Labs Tech. J. 2014, 18, 3–17. [CrossRef]
166. Shawkat Ali, A.B.M.; Azad, S. Demand Forecasting in Smart Grid. In Smart Grids; Springer: London, UK, 2013; pp. 135–150.
[CrossRef]
167. Khodayar, M.E.; Wu, H. Demand Forecasting in the Smart Grid Paradigm: Features and Challenges. Electr. J. 2015, 28, 51–62.
[CrossRef]
168. Mateo, F.; Carrasco, J.J. Machine Learning Techniques for Short-Term Electric Power Demand Prediction. Available on-
line: https://www.semanticscholar.org/paper/Machine-Learning-Techniques-for-Short-Term-Electric-Mateo-Fern%C3%A1
ndez/135715962dae1875bb4d67f2d6aed795a8178bdc (accessed on 2 June 2022).
169. Kotillova, A.; Koprinska, I.; Rana, M. Statistical and Machine Learning Methods for Electricity Demand Prediction. Lect. Notes
Comput. Sci. 2012, 2012, 535–542. [CrossRef]
170. Khotanzad, A.; Elragal, H.; Lu, T.L. Combination of Artificial Neural-Network Forecasters for Prediction of Natural Gas
Consumption. IEEE Trans. Neural Netw. 2000, 11, 464–473. [CrossRef]
171. Jin-ming, W.; Xin-heng, L. The Forecast of Energy Demand on Artificial Neural Network. Network 2009, 3, 31–35. [CrossRef]
172. Li, Z.-S.; Zhang, G.-Q.; Li, D.-M.; Liu, X.-H.; Mei, S.; Wu, J. Neural Network Prediction of Energy Demand and Supply in China.
Proc. Inst. Civ. Eng.-Energy 2007, 60, 4. [CrossRef]
173. Chen, B.; Zhao, L.; Lu, J.H. Wind Power Forecast Using RBF Network and Culture Algorithm. In Proceedings of the 1st
International Conference on Sustainable Power Generation and Supply, SUPERGEN ’09, Nanjing, China, 6–7 April 2009.
[CrossRef]
Electronics 2022, 11, 2676 29 of 33

174. Chang, W. Short-Term Load Forecasting Using Radial Basis Function Neural Network. J. Comput. Commun. 2015, 3, 40–45.
[CrossRef]
175. Sim, M.; Musilek, P.; Pelik, E. Recurrent Neural Network Based Gating for Natural Gas Load Prediction System. In Proceedings
of the 2006 IEEE International Joint Conference on Neural Network Proceedings, Vancouver, BC, Canada, 16–21 July 2006;
pp. 3736–3741.
176. Lee, K.Y.; Cha, Y.T.; Ku, C.C. A Study on Neural Networks for Short-Term Load Forecasting. In Proceedings of the First
International Forum on Applications of Neural Networks to Power Systems, Seattle, WA, USA, 23–26 July 1991.
177. Neupane, B.; Perera, K.; Aung, Z.; Woon, W. Artificial Neural Network-Based Electricity Price Forecasting for Smart Grid
Deployment. In Proceedings of the 2012 International Conference on Computer Systems and Industrial Informatics, Sharjah,
United Arab Emirates, 18–20 December 2012; pp. 1–6. [CrossRef]
178. Kumaran, J.; Ravi, G. Long-Term Sector-Wise Electrical Energy Forecasting Using Artificial Neural Network and Biogeography-
Based Optimization. Electr. Power Compon. Syst. 2015, 43, 1225–1235. [CrossRef]
179. Pellegrini, M. Short-Term Load Demand Forecasting in Smart Grids Using Support Vector Regression. In Proceedings of the 2015
IEEE 1st International Forum on Research and Technologies for Society and Industry Leveraging a Better Tomorrow (RTSI), Turin,
Italy, 16–18 September 2015; pp. 264–268.
180. Aung, Z.; Toukhy, M.; Williams, J. Towards Accurate Electricity Load Forecasting in Smart Grids. Proc. DBKDA 2012, 2012, 51–57.
181. List, C.; Authors, O.F.; Masri, D.; Zeineldin, H.; Woon, W.L. Electricity Price and Demand Forecasting under Smart Grid
Environment. In Proceedings of the 2015 IEEE 15th International Conference on Environment and Electrical Engineering (EEEIC),
Rome, Italy, 10–13 June 2015; pp. 2–4.
182. Sun, Y.; Braga-Neto, U.; Dougherty, E. Impact of Missing Value Imputation on Classification for DNA Microarray Gene Expression
Data—A Model-Based Study. EURASIP J. Bioinform. Syst. Biol. 2009, 2009, 504069. [CrossRef]
183. Mocanu, E.; Nguyen, P.H.; Gibescu, M.; Larsen, E.M.; Pinson, P. Demand Forecasting at Low Aggregation Levels Using Factored
Conditional Restricted Boltzmann Machine. In Proceedings of the 2016 Power Systems Computation Conference (PSCC), Genoa,
Italy, 20–24 June 2016.
184. Hossain, E.; Khan, I.; Un-Noor, F.; Sikander, S.S.; Sunny, M.S.H. Application of Big Data and Machine Learning in Smart Grid,
and Associated Security Concerns: A Review. IEEE Access 2019, 7, 13960–13988. [CrossRef]
185. Bazzan, A.L.C.; Klügl, F. Introduction to Intelligent Systems in Traffic And Transportation. In Synthesis Lectures on Artificial
Intelligence and Machine Learning (SLAIML); Springer: Berlin/Heidelberg, Germany, 2013; Volume 7, pp. 1–137. [CrossRef]
186. Zantalis, F.; Koulouras, G.; Karabetsos, S.; Kandris, D. A Review of Machine Learning and IoT in Smart Transportation. Future
Internet 2019, 11, 94. [CrossRef]
187. Gónzalez, L.C.; Martínez, F.; Carlos, M.R. Identifying Roadway Surface Disruptions Based on Accelerometer Patterns. IEEE Lat.
Am. Trans. 2014, 12, 455–461.
188. Brisimi, T.S.; Cassandras, C.G.; Osgood, C.; Paschalidis, I.C.; Zhang, Y. Sensing and Classifying Roadway Obstacles in Smart
Cities: The Street Bump System. IEEE Access Smart Cities 2016, 3536, 1301–1312. [CrossRef]
189. Darbari, M.; Yagyasen, D.; Tiwari, A. Intelligent Traffic Monitoring Using Internet of Things (IoT) with Semantic Web. Adv. Intell.
Syst. Comput. 2015, 1, 455–462. [CrossRef]
190. Liang, M.; Huang, X.; Chen, C.; Chen, X.; Tokuta, A. Counting and Classification of Highway Vehicles by Regression Analysis.
IEEE Trans. Intell. Transp. Syst. 2015, 16, 2878–2888. [CrossRef]
191. Araghi, S.; Khosravi, A.; Johnstone, M.; Creighton, D. Intelligent Traffic Light Control of Isolated Intersections Using Machine
Learning Methods. In Proceedings of the 2013 IEEE International Conference on Systems, Man, and Cybernetics, Manchester, UK,
13–16 October 2013; pp. 3627–3632. [CrossRef]
192. Mannion, P.; Duggan, J.; Howley, E. Parallel Reinforcement Learning for Traffic Signal Control. Procedia Comput. Sci. 2015, 52,
956–961. [CrossRef]
193. Alam, F.; Mehmood, R.; Katib, I. D2TFRS: An Object Recognition Method for Autonomous Vehicles Based on RGB and Spatial
Values of Pixels. In Proceedings of the Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications
Engineering, LNICST; Springer, Cham, Switzerland, 2018; Volume 224, pp. 155–168.
194. Hsu, J. Deep Learning Makes Driverless Cars Better at Spotting Pedestrians. Available online: http://spectrum.ieee.org/cars-
that-think/transportation/advanced-cars/deep-learning-makes-driverless-cars-better-at-spotting-pedestrians (accessed on 10
June 2022).
195. Korosec, K. Elon Musk Says Tesla Vehicles Will Drive Themselves in Two Years. Available online: http://fortune.com/2015/12/
21/elon-musk-interview/ (accessed on 10 June 2022).
196. Cook, D.J.; Youngblood, M.; Das, S.K. A Multi-Agent Approach to Controlling a Smart Environment. Lect. Notes Comput. Sci.
2006, 4008, 165–182. [CrossRef]
197. Kabir, M.H.; Hoque, M.R.; Seo, H.; Yang, S. Machine Learning Based Adaptive Context-Aware System for Smart Home
Environment. Int. J. Smart Home 2015, 9, 55–62. [CrossRef]
198. Dixit, A.; Naik, A. Use of Prediction Algorithms in Smart Homes. Int. J. Mach. Learn. Comput. 2014, 4, 157–162. [CrossRef]
199. Choi, S.; Kim, E.; Oh, S. Human Behavior Prediction for Smart Homes Using Deep Learning. In Proceedings of the 2013 RO-MAN,
Gyeongju, Korea, 26–29 August 2013; IEEE: Manhattan, NY, USA, 2013; p. 1.
Electronics 2022, 11, 2676 30 of 33

200. Fahad, L.G.; Ali, A.; Rajarajan, M. Learning Models for Activity Recognition in Smart Homes. Lect. Notes Electr. Eng. 2015, 339,
819–826. [CrossRef]
201. Taiwo, O.; Ezugwu, A.E.; Oyelade, O.N.; Almutairi, M.S. Enhanced Intelligent Smart Home Control and Security System Based
on Deep Learning Model. Wirel. Commun. Mob. Comput. 2022, 2022, 961. [CrossRef]
202. Islam, S.M.R.; Kwak, D.; Kabir, H.; Hossain, M.; Kwak, K.-S. The Internet of Things for Health Care: A Comprehensive Survey.
IEEE Access 2015, 3, 678–708. [CrossRef]
203. Bellandi, V.; Ceravolo, P.; Damiani, E.; Siccardi, S. Smart Healthcare, IoT and Machine Learning: A Complete Survey. Intell. Syst.
Ref. Libr. 2022, 212, 307–330. [CrossRef]
204. Chatzinikolaou, T.; Vogiatzi, E.; Kousis, A.; Tjortjis, C. Smart Healthcare Support Using Data Mining and Machine Learning. In
EAI/Springer Innovations in Communication and Computing; Springer: Cham, Switzerland, 2022; pp. 27–48.
205. Dawadi, P.N.; Cook, D.J.; Schmitter-edgecombe, M.; Parsey, C. Automated Assessment of Cognitive Health Using Smart Home
Technologies. Technol. Health Care 2013, 21, 323–343. [CrossRef]
206. Dawadi, P.N.; Cook, D.J.; Schmitter-edgecombe, M. Automated Cognitive Health Assessment Using Smart Home Monitoring of
Complex Tasks. IEEE Trans. Syst. Man Cybern. Syst. 2013, 43, 1302–1313. [CrossRef]
207. Vyas, N.; Farringdon, J.; Andre, D.; Stivoric, J.I. Machine Learning and Sensor Fusion for Estimating Continuous Energy
Expenditure. In Proceedings of the Twenty-Third Innovative Applications of Artificial Intelligence Conference, San Francisco,
CA, USA, 9–11 August 2011; pp. 1613–1620.
208. Boursalie, O.; Samavi, R.; Doyle, T.E. M4CVD: Mobile Machine Learning Model for Monitoring Cardiovascular Disease. Procedia
Comput. Sci. 2015, 63, 384–391. [CrossRef]
209. Smith, S. 5 Ways the IBM Watson Is Changing Health Care, from Diagnosing Disease to Treating It. Available online: http:
//www.medicaldaily.com/5-ways-ibm-watson-changing-health-care-diagnosing-disease-treating-it-364394 (accessed on 12
June 2022).
210. Ellis, S.; Morris, H.D.; Santagate, J. IoT-Enabled Analytic Applications Revolutionize Supply Chain Planning and Execution. Iran.
J. Inform. Process. Manag. 2015, 37, 1.
211. Braun, G. The Internet of Things and the Modern Supply Chain. 2015. Available online: www.c3solutions.com (accessed on 2
June 2022).
212. Deliottee. The Warehouse of Tomorrow, Today Ubiquitous Computing, Internet of Things, Machine Learning. 2015. Available
online: https://www2.deloitte.com/be/en/pages/technology/enterprise-technology-and-performance/solutions/emea-center-
of-excellence-for-sap-extended-warehouse-management.html (accessed on 2 June 2022).
213. Carbonneau, R.; Laframboise, K. Forecasting Supply Chain Demand Using Machine Learning Algorithms. In Distributed Artificial
Intelligence, Agent Technology, and Collaborative Applications; Oakland University: Oakland, CA, USA, 2009; pp. 2009–2011.
[CrossRef]
214. Zhengxia, W.; Laisheng, X. Modern Logistics Monitoring Platform Based on the Internet of Things. In Proceedings of the 2010
International Conference on Intelligent Computation Technology and Automation, Changsha, China, 11–12 May 2010.
215. Gai, K.; Qiu, M.; Zhao, H.; Dai, W. Anti-Counterfeit Scheme Using Monte Carlo Simulation for E-Commerce in Cloud Systems. In
Proceedings of the 2015 IEEE 2nd International Conference on Cyber Security and Cloud Computing, New York, NY, USA, 3–5
November 2015; pp. 74–79.
216. Machida, T.; Yamamoto, D.; Iwamoto, M.; Sakiyama, K. Implementation of Double Arbiter PUF and Its Performance Evaluation.
In Proceedings of the 20th Asia and South Pacific Design Automation Conference (ASP-DAC), Chiba, Japan, 19–22 January 2015;
pp. 6–7.
217. Liu, B.; Zhang, L. A Survey of Opinion Mining and Sentiment Analysis. In Mining Text Data; Aggarwal, C., Zhai, C., Eds.;
Springer: New York, NY, USA, 2013; pp. 415–463.
218. Bouchlaghem, R. A Machine Learning Approach for Classifying Sentiments in Arabic Tweets. In Proceedings of the 6th
International Conference on Web Intelligence, Mining and Semantics, Nîmes, France, 13–15 June 2016.
219. Duwairi, R.M. Arabic Sentiment Analysis Using Supervised Classification. In Proceedings of the 2014 International Conference
on Future Internet of Things and Cloud (FiCloud), Barcelona, Spain, 27–29 August 2014.
220. Science, C.; Trivandrum, E.; Science, C.; Trivandrum, E. Sentiment Analysis in Twitter Using Machine Learning Techniques.
In Proceedings of the 2013 Fourth International Conference on Computing, Communications and Networking Technologies
(ICCCNT), Tiruchengode, India, 4–6 July 2013.
221. Fisher, A.; Prucha, N. Detecting Jihadist Messages on Twitter. In Proceedings of the 2015 European Intelligence and Security
Informatics Conference, Manchester, UK, 7–9 September 2015.
222. Akbar, A.; Carrez, F.; Moessner, K.; Zoha, A. Predicting Complex Events for Pro-Active IoT Applications. In Proceedings of the
2015 IEEE 2nd World Forum on Internet of Things (WF-IoT), Milan, Italy, 14 –16 December 2015.
223. Roberts, F. IoT Project Brings Clean, Safe Water to Thousands in Africa. Available online: https://internetofbusiness.com/iot-
project-clean-safe-water-africa/ (accessed on 20 June 2022).
224. Shafi, U.; Mumtaz, R.; Anwar, H.; Qamar, A.M.; Khurshid, H. Surface Water Pollution Detection Using Internet of Things. In
Proceedings of the 15th International Conference on Smart Cities: Improving Quality of Life Using ICT & IoT (HONET-ICT),
Islamabad, Pakistan, 8–10 October 2018.
Electronics 2022, 11, 2676 31 of 33

225. Mishra, A. Air Pollution Monitoring System Based on IoT: Forecasting and Predictive Modeling Using Machine Learning. In
Proceedings of the Conference: International Conference on Applied Electromagnetics, Signal Processing & Communication
(AESPC), Bhubaneswar, India, 22–24 October 2018.
226. Elvitigala, C.S.; Sudantha, B. Machine Learning Capable, IoT Air Pollution Monitoring System with Upgradable Sensor Array. In
Proceedings of the ISIS Conference, Daegu, Korea, 11–14 October 2017.
227. Ray, P.P.; Mukherjee, M.; Shu, L. Internet of Things for Disaster Management: State-of-the-Art and Prospects. IEEE Access 2017, 5,
18818–18835. [CrossRef]
228. Nesa, N.; Ghosh, T.; Banerjee, I. Outlier Detection in Sensed Data Using Statistical Learning Models for IoT. In Proceedings of the
2018 IEEE Wireless Communications and Networking Conference (WCNC), Barcelona, Spain, 15–18 April 2018; pp. 1–6.
229. Salehi, M.; Rashidi, L. A Survey on Anomaly Detection in Evolving Data: With Application to Forest Fire Risk Prediction. SIGKDD
Explor. Newsl. 2018, 20, 13–23. [CrossRef]
230. Pughazhendhi, G.; Raja, A.; Ramalingam, P.; Elumalai, D.K. Earthosys—Tsunami Prediction and Warning System Using Machine
Learning and IoT. In Proceedings of the International Conference on Computational Intelligence and Data Engineering; Chaki, N.,
Devarakonda, N., Sarkar, A., Debnath, N.C., Eds.; Springer: Singapore, 2019; pp. 103–113.
231. Knight, W. How Japan’s Earthquake and Tsunami Warning Systems Work. Available online: https://www.technologyreview.
com/2011/03/11/260021/how-japans-earthquake-and-tsunami-warning-systems-work/ (accessed on 2 June 2022).
232. Andaloussi, Y.; El Ouadghiri, M.D.; Maurel, Y.; Bonnin, J.M.; Chaoui, H. Access Control in IoT Environments: Feasible Scenarios.
In Proceedings of the 8th International Symposium on Frontiers in Ambient and Mobile Systems (FAMS-2018), Brisbane, QSL,
Australia, 8 May 2018; pp. 1031–1036.
233. Xiao, L.; Wan, X.; Lu, X.; Zhang, Y.; Wu, D. IoT Security Techniques Based on Machine Learning: How Do IoT Devices Use AI to
Enhance Security? IEEE Signal Process. Mag. 2018, 35, 41–49. [CrossRef]
234. Hussain, F.; Hussain, R.; Hassan, S.A.; Hossain, E. Machine Learning in IoT Security: Current Solutions and Future Challenges.
IEEE Commun. Surv. Tutor. 2020, 22, 1686–1721. [CrossRef]
235. Hasan, M.; Islam, M.D.M.; Zarif, M.I.I.; Hashem, M.M.A. Attack and Anomaly Detection in IoT Sensors in IoT Sites Using
Machine Learning Approaches. Internet Things 2019, 7, 100059. [CrossRef]
236. Khalifa, W.H.; Roushdy, M.I.; Salem, A.B.-M. Machine Learning Techniques for Intelligent Access Control. In New Ap-
proaches in Intelligent Image Analysis. Intelligent Systems Reference Library; Kountchev, R., Nakamatsu, K., Eds.; Springer: Cham,
Switzerland, 2019.
237. Lee, S.-H.; Yang, C.-S. An Intelligent Home Access Control System Using Deep Neural Network. In Proceedings of the 2017 IEEE
International Conference on Consumer Electronics—Taiwan (ICCE-TW), Taipei, Taiwan, 12–14 June 2017.
238. Marr, B. What Is Industry 4.0? Here’s A Super Easy Explanation for Anyone. Available online: https://www.forbes.com/sites/
bernardmarr/2018/09/02/what-is-industry-4-0-heres-a-super-easy-explanation-for-anyone/?sh=c8fe9769788a (accessed on 23
June 2022).
239. Barnes, J. Azure Machine Learning Microsoft Azure Essentials; Microsoft Press: New York, NY, USA, 2015; ISBN 9780735698178.
240. Barga, R.; Fontama, V.; Tok, W.H. Predictive Analytics with Microsoft Azure Machine Learning. In Predictive Analytics with
Microsoft Azure Machine Learning; Apress: New York, NY, USA, 2015; ISBN 978-1-4842-1200-4.
241. Studio, M.V. Scott Hanselman’s Best Demo! IoT, Azure, Machine Learning & More. Available online: https://www.youtube.
com/watch?v=u5oTz1e5qqE (accessed on 2 June 2022).
242. Higginbotham, S. IBM Is Bringing in Watson to Conquer The Internet of Things. Available online: http://fortune.com/2015/12/
15/ibm-watson-iot/ (accessed on 23 June 2022).
243. IBM. Watson Internet of Things. Available online: http://www.ibm.com/internet-of-things/iot-solutions/watson-iot-platform/
(accessed on 23 June 2022).
244. Knowles, G.; Melamed, A.; Fisher, A. Accelerate the Development of Cognitive Computing in Your IoT App. Available
online: http://www.ibm.com/developerworks/library/iot-cc-watson-iot-platform-trs/iot-cc-watson-iot-platform-trs-pdf.pdf
(accessed on 20 June 2022).
245. News Room IBM and Cisco Combine the Power of Watson Internet of Things with Edge Analytics. Available online: https:
//www-03.ibm.com/press/us/en/pressrelease/49845.wss (accessed on 25 June 2022).
246. Tilley, A. IBM And Cisco Partner to Bring Watson Closer To The “Internet Of Things”. Available online: http://www.forbes.com/
sites/aarontilley/2016/06/02/ibm-cisco-watson-internet-of-things/#1c0c0bda564a (accessed on 23 June 2022).
247. Bilac, S. Machine Learning in the Cloud, with TensorFlow. Available online: https://research.googleblog.com/2016/03/machine-
learning-in-cloud-with.html (accessed on 23 June 2022).
248. Kepes, B. Splunk Ups the Machine-Learning Ante. Available online: http://www.computerworld.com/article/3125517/big-
data/splunk-ups-the-machine-learning-ante.html (accessed on 26 June 2022).
249. Morphy, E. Splunk Enhances Machine Learning with 300 New Algorithms. Available online: http://www.cmswire.com/big-
data/splunk-enhances-machine-learning-with-300-new-algorithms/ (accessed on 2 June 2022).
250. News AWS IoT Now Supports Integration with Amazon Machine Learning and AWS CloudTrail. Available online:
https://aws.amazon.com/about-aws/whats-new/2016/04/aws-iot-now-supports-integration-with-amazon-machine-
learning-and-aws-cloudtrail/ (accessed on 25 June 2022).
Electronics 2022, 11, 2676 32 of 33

251. Yegulalp, S. Amazon Dumbs down Machine Learning for the Rest of Us. Available online: http://www.infoworld.com/article/
2908498/amazon-web-services/amazon-machine-learning-for-the-rest-us.html (accessed on 27 June 2022).
252. CB Insights. Sensors, Plus Brains: 17 IoT Companies Using Artificial Intelligence Tech. Available online: https://www.cbinsights.
com/blog/artificial-intelligence-iot-company-list/ (accessed on 25 June 2022).
253. Elayan, H.; Aloqaily, M.; Member, S.; Karray, F.; Guizani, M. Internet of Behavior (IoB) and Explainable AI Systems for Influencing IoT
Behavior; Cornell University: Ithaca, NY, USA, 2022.
254. Javaid, M.; Haleem, A.; Singh, R.P.; Rab, S.; Suman, R. Internet of Behaviours (IoB) and Its Role in Customer Services. Sens. Int.
2021, 2, 100122. [CrossRef]
255. Pinochet, L.H.C.; Lopes, E.L.; Srulzon, C.H.F.; Onusic, L.M. The Influence of the Attributes of “Internet of Things” Products on
Functional and Emotional Experiences of Purchase Intention. Innov. Manag. Rev. 2018, 15, 303–320. [CrossRef]
256. Stary, C. The Internet-of-Behavior as Organizational Transformation Space with Choreographic Intelligence. Commun. Comput.
Inf. Sci. 2020, 1278, 113–132. [CrossRef]
257. Singh, R.P.; Javaid, M.; Haleem, A.; Suman, R. Internet of Things (IoT) Applications to Fight against COVID-19 Pandemic. Diabetes
Metab. Syndr. Clin. Res. Rev. 2020, 14, 521–524. [CrossRef]
258. Kolankari, P. As COVID-19 Hounds the World, AI and IoT Lead the Charge in Healthcare 4.0. Economic Times, 4 August 2020.
259. Mishra, S. The Growing Role of IoT in COVID-19 Response. Available online: https://www.iotforall.com/the-growing-role-of-
iot-in-covid-19-response/ (accessed on 16 June 2022).
260. Higginbotham, S. Tracking COVID-19 With the IoT May Put Your Privacy at Risk. Available online: https://spectrum.ieee.org/
tracking-covid19-with-the-iot-may-put-your-privacy-at-risk (accessed on 28 June 2022).
261. Alam, F.; Almaghthawi, A.; Katib, I.; Albeshri, A.; Mehmood, R. IResponse: An AI and IoT-Enabled Framework for Autonomous
COVID-19 Pandemic Management. Sustainability 2021, 13, 3797. [CrossRef]
262. Talebian, A.; Mishra, S. Predicting the Adoption of Connected Autonomous Vehicles: A New Approach Based on the Theory of
Diffusion of Innovations. Transp. Res. Part C Emerg. Technol. 2018, 95, 363–380. [CrossRef]
263. Alam, F.; Mehmood, R.; Katib, I.; Altowaijri, S.; Albeshri, A. TAAWUN: A Decision Fusion and Feature Specific Road Detection
Approach for Connected Autonomous Vehicles. Mob. Netw. Appl. 2019, 2019, 1–17. [CrossRef]
264. Morris, D.Z. Google Self-Driving Car in Serious Crash in Mountain View. Fortune. 2016. Available online: http://fortune.com/
2016/09/25/crash-google-self-driving-car/ (accessed on 20 June 2022).
265. Levin, S.; Woolf, N. Tesla Driver Killed While Using Autopilot Was Watching Harry Potter, Witness Says. 2016. Available online:
https://www.theguardian.com/technology/2016/jul/01/tesla-driver-killed-autopilot-self-driving-car-harry-potter (accessed
on 20 June 2022).
266. Elliott, D.; Keen, W.; Miao, L. Recent Advances in Connected and Automated Vehicles. J. Traffic Transp. Eng. 2019, 6, 109–131.
[CrossRef]
267. Ye, L.; Yamamoto, T. Evaluating the Impact of Connected and Autonomous Vehicles on Traffic Safety. Phys. A Stat. Mech. Its Appl.
2019, 526, 121009. [CrossRef]
268. Cisco. Fog Computing and the Internet of Things: Extend the Cloud to Where the Things Are. CISCO White Paper 2016, 6.
Available online: https://docslib.org/doc/7133745/fog-computing-and-the-internet-of-things-extend-the-cloud-to-where-the-
things-are (accessed on 2 June 2022).
269. Gonzalez, N.M.; Goya, W.A.; De, R.; Pereira, F.; Langona, K.; Silva, E.A.; Melo De Brito Carvalho, C.; Miers, C.C.; Mångs, J.-E.;
Sefidcon, A. Fog Computing: Data Analytics and Cloud Distributed Processing on the Network Edges. In Proceedings of the
2016 35th International Conference of the Chilean Computer Science Society (SCCC), Valparaiso, Chile, 10–14 October 2016.
270. Greenfield, D. Fog Computing vs. Edge Computing: What’s the Difference? Available online: https://www.automationworld.
com/fog-computing-vs-edge-computing-whats-difference (accessed on 28 June 2022).
271. LeCun, Y.; Bengio, Y.; Hinton, G. Deep Learning. Nature 2015, 521, 436–444. [CrossRef]
272. Yao, S.; Zhao, Y.; Zhang, A.; Hu, S.; Shao, H.; Zhang, C.; Su, L.; Abdelzaher, T. Deep Learning for the Internet of Things. Computer
2018, 51, 32–41. [CrossRef]
273. Zhao, T.; Xie, Y.; Wang, Y.; Cheng, J.; Guo, X.; Hu, B.; Chen, Y. A Survey of Deep Learning on Mobile Devices: Applications,
Optimizations, Challenges, and Research Opportunities. Proc. IEEE 2022, 110, 334–354. [CrossRef]
274. Duong, C.N.; Quach, K.G.; Jalata, I.; Le, N.; Luu, K. MobiFace: A Lightweight Deep Learning Face Recognition on Mobile Devices.
In Proceedings of the 2019 IEEE 10th International Conference on Biometrics Theory, Applications and Systems (BTAS), Tampa,
FL, USA, 23 September 2019; pp. 1–6.
275. Agarwal, P.; Alam, M. A Lightweight Deep Learning Model for Human Activity Recognition on Edge Devices. Procedia Comput.
Sci. 2020, 167, 2364–2373. [CrossRef]
276. Martindez-Díaz, Y.; Luevano, L.S.; Mendez-Vazquez, H.; Nicolas-Diaz, M.; Chang, L.; Gonzalez-Mendoza, M. ShuffleFaceNet:
A Lightweight Face Architecture for Efficient and Highly-Accurate Face Recognition. In Proceedings of the 2019 IEEE/CVF
International Conference on Computer Vision Workshop (ICCVW), Seoul, Korea, 27–28 October 2019. [CrossRef]
277. Sliwa, B.; Piatkowski, N.; Wietfeld, C. LIMITS: Lightweight Machine Learning for IoT Systems with Resource Limitations. In
Proceedings of the ICC 2020—2020 IEEE International Conference on Communications (ICC), Dublin, Ireland, 7–11 June 2020.
[CrossRef]
Electronics 2022, 11, 2676 33 of 33

278. Shuvo, S.B.; Ali, S.N.; Swapnil, S.I.; Al-Rakhami, M.S.; Gumaei, A. CardioXNet: A Novel Lightweight Deep Learning Framework
for Cardiovascular Disease Classification Using Heart Sound Recordings. IEEE Access 2021, 9, 36955–36967. [CrossRef]
279. Luo, L. Applying Lightweight Deep Learning-Based Virtual Vision Sensing Technology to Realize and Develop New Media
Interactive Art Installation. Comput. Intell. Neurosci. 2022, 2022, 9119316. [CrossRef]
280. Pang, S.; Wang, S.; Rodríguez-Patón, A.; Li, P.; Wang, X. An Artificial Intelligent Diagnostic System on Mobile Android Terminals
for Cholelithiasis by Lightweight Convolutional Neural Network. PLoS ONE 2019, 14, e0221720. [CrossRef]
281. Winoto, A.S.; Kristianus, M.; Premachandra, C. Small and Slim Deep Convolutional Neural Network for Mobile Device. IEEE
Access 2020, 8, 125210–125222. [CrossRef]
282. Nižetić, S.; Šolić, P.; López-de-Ipiña González-de-Artaza, D.; Patrono, L. Internet of Things (IoT): Opportunities, Issues, and
Challenges towards a Smart and Sustainable Future. J. Clean. Prod. 2020, 274, 122877. [CrossRef]
283. Idrees Sarhan, Q. Internet of Things: A Survey of Challenges and Issues. Int. J. Internet Things Cyber Assur. 2018, 1, 40–75.
[CrossRef]
284. Malekshahi Rad, M.; Rahmani, A.M.; Sahafi, A.; Nasih Qader, N. Social Internet of Things: Vision, Challenges, and Trends. Hum.
Cent. Comput. Inf. Sci. 2020, 10, 52. [CrossRef]
285. Sadique, K.M.; Rahmani, R.; Johannesson, P. Towards Security on Internet of Things: Applications and Challenges in Technology.
Procedia Comput. Sci. 2018, 141, 199–206. [CrossRef]
286. O’Halloran, J. Connectivity Issues Disrupting Most Businesses’ IoT Roll-Outs. Available online: https://www.computerweekly.
com/news/252509991/Connectivity-issues-disrupting-most-businesses-IoT-roll-outs (accessed on 2 June 2022).
287. IoT Connectivity: Challenges and Strategies. Available online: https://www.telecomreview.com/index.php/articles/reports-
and-coverage/5319-iot-connectivity-challenges-and-strategies (accessed on 2 June 2022).
288. Rahul, Y.; Weizhe, Z.; Neeraj, K.; Omprakash, K. Recent Trends and Challenges in Fog/Edge Computing for Internet of Things.
Available online: https://www.hindawi.com/journals/wcmc/si/975923/ (accessed on 2 June 2022).
289. Abi Sen, A.A.; Eassa, F.A.; Jambi, K.; Yamin, M. Preserving Privacy in Internet of Things: A Survey. Int. J. Inf. Technol. 2018, 10,
189–200. [CrossRef]
290. Falcone, R.; Sapienza, A. On the Users’ Acceptance of IoT Systems: A Theoretical Approach. Information 2018, 9, 53. [CrossRef]
291. Beştepe, F.; Yildirim, S.Ö. Acceptance of IoT-Based and Sustainability-Oriented Smart City Services: A Mixed Methods Study.
Sustain. Cities Soc. 2019, 526, 121009. [CrossRef]
292. Patil, Y. 6 Key Challenges to Consider for Successful IoT Implementation. Available online: https://www.saviantconsulting.com/
blog/iot-implementation-challenges-enterprises.aspx (accessed on 2 June 2022).
293. Akhairamka, P. The 15 Key Challenges with IoT Project Development in 2021. Available online: https://iot-analytics.com/
challenges-iot-project-development/ (accessed on 2 June 2022).
294. IoT. Business.News IoT News—IoT Market Growth Will Slow down due to the COVID-19 Pandemic and the Limited Take-up of
LPWA Solutions. Available online: https://iotbusinessnews.com/2021/02/25/20654-iot-market-growth-will-slow-down-due-
to-the-covid-19-pandemic-and-the-limited-take-up-of-lpwa-solutions/ (accessed on 2 June 2022).
295. Developer Burnout and a Global Chip Shortage: The IoT Is Facing a Perfect Storm|ZDNet. Available online: https://www.zdnet.
com/article/developer-burn-out-and-a-global-chip-shortage-the-iot-is-facing-a-perfect-storm/ (accessed on 2 June 2022).
296. IDC. News European IoT Spending Continues Its Double-Digit Growth, Despite Global Uncertainty and Slow Demand, Says
IDC. Available online: https://www.idc.com/getdoc.jsp?containerId=prEUR149276822 (accessed on 2 June 2022).
297. Bernard Marr The 5 Biggest Internet of Things (IoT) Trends In 2021 Everyone Must Get Ready for Now. Available on-
line: https://www.forbes.com/sites/bernardmarr/2020/10/26/the-5-biggest-internet-of-things-iot-trends-in-2021-everyone-
must-get-ready-for-now/?sh=5c8dd2c641fd (accessed on 2 June 2022).
298. Chohan, S.R.; Hu, G. Success Factors Influencing Citizens’ Adoption of IoT Service Orchestration for Public Value Creation in
Smart Government. IEEE Access 2020, 8, 208427–208448. [CrossRef]