Military ML

mathematics
Article
Military Applications of Machine Learning:
A Bibliometric Perspective
José Javier Galán 1, *, Ramón Alberto Carrasco 2 and Antonio LaTorre 3
1 Faculty of Statistics, Complutense University, Puerta de Hierro, 3728040 Madrid, Spain

2 Department of Management and Marketing, Faculty of Commerce and Tourism Complutense, University of
Madrid, 28223 Madrid, Spain; [email protected]
3 Center for Computational Simulation (CCS), Universidad Politécnica de Madrid, 28660 Madrid, Spain;
[email protected]
* Correspondence: [email protected]
Abstract: The military environment generates a large amount of data of great importance, which
makes necessary the use of machine learning for its processing. Its ability to learn and predict possible
scenarios by analyzing the huge volume of information generated provides automatic learning and
decision support. This paper aims to present a model of a machine learning architecture applied to a
military organization, carried out and supported by a bibliometric study applied to an architecture
model of a nonmilitary organization. For this purpose, a bibliometric analysis up to the year 2021
was carried out, making a strategic diagram and interpreting the results. The information used
has been extracted from one of the main databases widely accepted by the scientific community,
ISI WoS. No direct military sources were used. This work is divided into five parts: the study of
previous research related to machine learning in the military world; the explanation of our research
methodology using the SciMat, Excel and VosViewer tools; the use of this methodology based on data
mining, preprocessing, cluster normalization, a strategic diagram and the analysis of its results to
investigate machine learning in the military context; based on these results, a conceptual architecture
of the practical use of ML in the military context is drawn up; and, finally, we present the conclusions,
Citation: Galán, J.J.; Carrasco, R.A.; where we will see the most important areas and the latest advances in machine learning applied, in
LaTorre, A. Military Applications of
this case, to a military environment, to analyze a large set of data, providing utility, machine learning
Machine Learning: A Bibliometric
and decision support.
Perspective. Mathematics 2022, 10,
1397. https://doi.org/10.3390/
Keywords: machine learning; military; artificial intelligence; bibliometric analysis
math10091397
Academic Editor: Alexander Ryzhov MSC: 68T01
Received: 28 February 2022

Accepted: 19 April 2022
Published: 22 April 2022
1. Introduction
Publisher’s Note: MDPI stays neutral
Machine learning (ML) allows the automation of many tasks by taking advantage
with regard to jurisdictional claims in
of the large amount of information available from different sources, including big data
published maps and institutional affil-
applications. Its use is currently widely spread, and ML has become an important part of
iations.
our daily lives [1].
In the military, the use of intelligent applications has also accelerated [2]. For example,
the South Korean Ministry of National Defense has increased its information significantly,
Copyright: © 2022 by the authors.
and with fewer and fewer intelligence analysts they need to apply artificial intelligence (AI)
Licensee MDPI, Basel, Switzerland. technology to process all the information in an accurate and timely manner [3]. Another
This article is an open access article example to note is the dependence on oil by military equipment and machinery. This is
distributed under the terms and also where ML comes in, as military logistics must be intelligently based on informed
conditions of the Creative Commons deductions [4]; thus, we see how ML is integrated into the military world.
Attribution (CC BY) license (https:// The objective of this paper is to present an architectural model that reflects how ML is
creativecommons.org/licenses/by/ applied in a practical way in the military environment. In this architecture, we solve aspects
4.0/). such as the most frequent data, algorithms and applications used in the military context.
Mathematics 2022, 10, 1397. https://doi.org/10.3390/math10091397 https://www.mdpi.com/journal/mathematics

Mathematics 2022, 10, 1397 2 of 27
While carrying out this work, as we will see in Section 2, we study related work, noting
that there are few review works in this emerging topic, which has aroused our interest in
performing a bibliometric analysis on one of the main scientific databases, Web of Science,
up to and including the year 2021. In the same section, we also present a conceptual
architecture for the application of ML in a practical way in a nonmilitary organization,
since there are no works reflecting such an architecture in the military domain.
The bibliometric methodology used in this work is explained in Section 3, and we will
mainly make use of the SciMat bibliometric analysis tool, capable of performing a scientific
mapping analysis in a longitudinal framework [5]. With this analysis we build a strategic
diagram in which we identify the main areas of ML applied to the military field.
In Section 4, we apply the described methodology to perform an analysis according
to origin: we see the main scientific areas in which ML is applied to the military world;
author and citation: we determine who are the most active authors in this subject; country:
we analyze how the countries that generate more scientific documentation in this sense
are usually the ones that have fewer citations; and we distinguish two periods: before
2015 and after 2016, after which the increase in publications on ML in the military world
rapidly ascends.
In Section 5, once this bibliometric analysis is completed, we are now in a position to
redefine the conceptual architecture presented in Section 2 specifically for military organizations.
Finally, we come to some conclusions in which we expose the results obtained related
to the main thematic areas found and the conclusions.
2. Related Work
First, in Section 2.1, we searched for bibliometric or review articles related to the
application of ML in the military world, and we did not find relevant information. Then,
in Section 2.2, we searched for a data-driven architecture for nonmilitary organizations
to serve as a basis for establishing a new model oriented to the military world, which we
complement with the present bibliometric study.
2.1. Previous Research on the Military Applications of ML

In this section, we studied review papers or bibliometric studies on AI (which includes
ML) and related areas, in addition to their applications in the military field. The results of
this study are shown in Table 1.
Table 1. Related review work.
Category of the Review Work Refs.

Robotics and smart devices with military applications [6–22]
Generic ML and optimization techniques with military applications [23–33]
ML and optimization techniques focused on military applications [34–36]
We have classified these works into three categories:

• Reviews on robotics and smart devices with military applications. We found several
works on this subject, including on drones, sensors, computer vision, unmanned aerial
vehicles, etc. These works have military applications, although they are not specifically
developed for the military field;
• Reviews on generic ML and optimization techniques with military applications. In
this category, we included several more or less generic review papers on ML and
optimization techniques, but always including applications to the military field;
• Reviews on ML and optimization techniques focused on military applications. This
category includes works specifically developed for the military field, and therefore
can be considered as being more related to the proposals of this article.
Mathematics 2022, 10, 1397 3 of 27
Due to its interest, we will analyze this last group in more detail. Firstly, we have found
recent review works on specific optimization techniques, such as dynamic programming
and its application to the military field [34].
We have also found a review on AI and its applications in the military field [35]. This
review describes three main military applications:
• Military AI overview. The main AI projects and milestones carried out by the Defense
Advanced Research Projects Agency (DARPA) are examined. In June 2016, the Alpha
AI air combat simulation opponent pilot developed by the U.S. University of Cincinnati
scored a victory over famed Air Force tactical instructor Colonel Gene Lee. As of 2018,
the AI Next project included five directions: new AI capability; robust AI; anti-AI;
high-performance AI and next-generation AI;
• Observation, orientation, decision and action approaches. The authors explore this
approach based primarily on data and AI in the military field:
# Information fusion. In the military field, merging heterogeneous information
from different sources is essential;
# Situation awareness. The perception of the existing elements in the volume of the
time and space of an environment, understanding its meaning, intention with
other agents and its future;
# Decision support system. Requires the participation of hybrid decision makers:
humans and computers;
# Path planning. Used to avoid threats or obstacles to help commanders choose the
appropriate path;
# Human–machine interface. Interrelating a mechanical element with an assistant
in an ideal environment.
• Challenges and solutions of AI in a military context. The main ones found by the
authors are as follows:
# Modeling of complex systems. A large amount of information is generated in
a war environment. The level of intelligence can be verified and evaluated by
means of a simulation system, which is quite reduced;
# Imperfect information environment. In a combat situation, the information ob-
tained is always limited, and the authenticity of the information is not guaranteed.
AI technology is not omnipotent; it must be combined with traditional technolo-
gies, such as knowledge reasoning and search and solution, in which the role of
domain knowledge is indispensable.
This work, while interesting, is not specifically oriented to ML, and also focuses almost
exclusively on U.S. applications. In this respect, it cannot be considered a global survey of
the field.
There are other, less important NATO (North Atlantic Treaty Organization) papers
that are coincident with the military applications of AI [31].
It can be seen that the existing scientific literature does not reflect specific and com-
prehensive studies on ML applied to the military environment. Therefore, we see the
need to carry out this novel study, which provides the scientific community with updated
data on the scientific interest of ML applied to the military environment, explaining the
identified areas and making a categorization of them, and showing the existing interest in
decision making based on data in such a military environment. In Section 3 we explain the
methodology used for this study, which aims to be more objective than most of the related
works by using a bibliometric study as a basis.
2.2. Background of Data-Driven Architecture for Organizations

Data-driven decision making has ML algorithms as key components. There are
works [37] that specify conceptual architectures that allow an organization to adopt this
philosophy in a practical way in a scalable context, which allows it to adapt to a big
data environment, i.e., with an increasing data volume and with a variety of formats
2.2. Background of Data-Driven Architecture for Organizations
Data-driven decision making has ML algorithms as key components. There are
Mathematics 2022, 10, 1397 works [37] that specify conceptual architectures that allow an organization to adopt4this of 27
philosophy in a practical way in a scalable context, which allows it to adapt to a big data
environment, i.e., with an increasing data volume and with a variety of formats (unstruc-
tured, semi-structured
(unstructured, and structured)
semi-structured that are produced;
and structured) therefore,therefore,
that are produced; they havethey
to be pro-
have to
cessed at high velocity. We have not found in the literature specific architectures
be processed at high velocity. We have not found in the literature specific architectures for mili-
tary organizations:
for military therefore,
organizations: we havewe
therefore, specified in Figure
have specified in 1Figure
a generic architecture,
1, a generic based
architecture,
on [37,38],
based that could
on [37,38], that support these types
could support these of organizations
types in the in
of organizations big
thedata
big context dis-
data context
cussed. As mentioned,
discussed. As mentioned,the the
main objective
main of of
objective this paper
this will
paper bebe
will totospecify
specifyininmore
moredetail
detail
the
thecomponents
componentsof ofsuch
suchananarchitecture
architecturefor
foraamilitary
militaryorganization.
organization.
Figure 1. Data-driven architecture for nonmilitary organizations.

Figure 1. Data-driven architecture for nonmilitary organizations.
The
Thecomponents
componentsof ofthe
theconceptual
conceptualarchitecture
architectureare areasasfollows:
follows:
o# Data management solutions for business analytics
Data management solutions for business analytics (DMSBA). (DMSBA). Data
Dataareare
thethe
essential
essential
raw materials for data-driven decision making. From
raw materials for data-driven decision making. From them, ML algorithms them, ML algorithms will be be
will
able to find the desired knowledge in the form of patterns. In
able to find the desired knowledge in the form of patterns. In this way, we should this way, we should
have
haveaccess
accesstoto both
bothinternal andand
internal external datadata
external considered
considered important
importantfor decision
for decision
making.
making.The The format
format of ofthese data
these cancan
data be structured
be structured (typically tables
(typically of relational
tables of relational
databases
databasesthat thatare
aremanipulated
manipulated with
withSQL),
SQL), semi-structured
semi-structured (with
(with structure
structurebutbutnotnot
tables,
tables,including
includingNoSQLNoSQLdatabases)
databases)ororunstructured
unstructured(without
(withoutaadefineddefinedformat,
format, such
such as natural
as natural language,
language, images,
images, video,
video, etc.)[37,39].
etc.) [37,39]. AA conventional organization
organization is
ismainly
mainlycharacterized
characterizedby bystructured
structuredinternal
internal datadata from
from its its operational
operational systems:
systems: ERP
ERP (enterprise
(enterprise resource
resource planning)
planning) and and
CRM CRM (customer
(customer relational
relational management).
management). Prior to
Prior
theirtoanalysis,
their analysis,
these data these databemust
must be stored.
stored. For thisFor this purpose,
purpose, there arethere
twoare twotypes
main
main types ofa systems:
of systems: a data warehouse,
data warehouse, mainly fordata,
mainly for structured structured data,lake,
and a data and for
a data
all other
cases.
lake, forData provided
all other cases.byData
experts shouldby
provided also be considered
experts should also in this layer, since in
be considered thisthis
type
of knowledge should be systematized in organizations [40];
layer, since this type of knowledge should be systematized in organizations [40];
o# Insight
Insightgeneration
generationfor for business.
business. This
This is the
is the tasktask performed
performed by data
by data scientists
scientists withwith
MLMLalgorithms
algorithmsatat itsits core.
core. MLML algorithms
algorithms areare classified
classified intointo
threethree
main main categories:
categories:
supervised learning, unsupervised learning and reinforcement learning. Supervised
ML uses labeled data in a certain target class, which must be learned by the algorithm.
If this target variable is continuous, we would be in a case of regression, as would be
the case if it is discrete in classification. Unsupervised ML works with unlabeled data
by obtaining groupings of the data, association rules, dimensionality reduction, etc.
supervised learning, unsupervised learning and reinforcement learning. Supervised
ML uses labeled data in a certain target class, which must be learned by the algo-
Mathematics 2022, 10, 1397 rithm. If this target variable is continuous, we would be in a case of regression,5 of as27
would be the case if it is discrete in classification. Unsupervised ML works with
unlabeled data by obtaining groupings of the data, association rules, dimensionality
reduction, etc. Reinforcement ML considers the problem of a computational agent
Reinforcement ML considers the problem of a computational agent learning to make
learning to make decisions by a trial-and-error method [39];
decisions by a trial-and-error method [39];
o Business Application. This is the task performed by data scientists with ML algo-
# Business Application. This is the task performed by data scientists with ML algo-
rithms at its core. Using the knowledge extracted in the previous layer, the appro-
rithms at its core. Using the knowledge extracted in the previous layer, the appropriate
priate business decisions are made in this layer. Although there are many business
business decisions are made in this layer. Although there are many business applica-
applications,
tions, the mostthe most important
important ones haveones
beenhave been identified
identified in shown
in [38], as [38], as in
shown
Figurein1.Fig-
A
ure 1. A preliminary view is that many of these applications could be
preliminary view is that many of these applications could be applicable in military applicable in
military
fields, fields,
such such as predictive
as predictive analytics
analytics and and intelligent
intelligent decision
decision making, making, cyber-
cybersecurity and
threat intelligence, healthcare, image, speech and pattern recognition, etc.recognition,
security and threat intelligence, healthcare, image, speech and pattern In this layer
etc. In
there is this layer there
a constant taskisofa monitoring
constant task of monitoring
and and permanent
permanent learning about the learning
decisions
about the decisions taken, in order to contrast them
taken, in order to contrast them with the business objectives. with the business objectives.
3.3.Research
ResearchMethodology
Methodology
InInthis
thissection
Sectionwewe explain
explain thethe methodology
methodology used,
used, which
which wewe will
will later
later apply
apply in Section
in Section 4.
4. The
The methodology
methodology usedused is based
is based on scheme
on the the scheme presented
presented in Figure
in Figure 2. 2.
Figure2.2.Steps
Figure Stepsused
usedininthe
theresearch
researchmethodology.
methodology.
3.1.
3.1.Set
Setthe
theObjectives
Objectivesofofthe
theAnalysis
Analysis
At
At this stage, the researchquestions
this stage, the research questionsshould
shouldbebeclearly
clearlyformulated.
formulated.
The research topic, the period to be investigated,
The research topic, the period to be investigated, the the sources
sourcestotobe
beused,
used,etc.,
etc.,must
mustbe
bedelimited.
delimited.
The
Theobjective
objectiveisistotoconduct
conducta acomprehensive
comprehensivestudy studyofofthe
therelationship
relationshipbetween
betweenthe the
military and ML, following on from the proposal of conceptual military
military and ML, following on from the proposal of conceptual military architecture.architecture.
We
Wewant
wanttotoknow
knowwhatwhatisisbeing
beingresearched,
researched,bybywhom,
whom, in in which
which countries and organ-
countries and orga-
nizations and how it is developing over time. We want to draw conclusions about where
izations and how it is developing over time. We want to draw conclusions about where
specific research topics could be focused. Finally, our objective is to present the main areas
identified that relate ML to the military environment, to group these areas into categories
that we will explain one by one in Section 4 and, finally, to comment on the dimensions
detected in relation to these topics.
Mathematics 2022, 10, 1397 6 of 27
3.2. Data Extraction

To carry out this study, we need to know the scientific literature related to the topic
“Machine Learning” and, at the same time, relating to the “military*” environment. We
have chosen as a timeline all the existing literature up to the year 2021, this year included.
Unfortunately, we have not had access to military databases or documents, and we have
only used the Web of Science core collection database, which is widely accepted in the
scientific environment. The query used in January 2022 was:
TS = (“machine learning” and “military*”)
Once we have the data, we proceed to its preprocessing.
3.3. Data Preprocessing

From the results obtained in the extraction, we selected those that are really related to
ML in the military environment, finally obtaining 525 documents.
Specifically, a standardization process was carried out by merging the plural and sin-
gular forms and converting the acronyms into their respective keywords using Levenshtein
distance in SciMAT.
3.4. Multidimensional Analysis

Qualitative data can be analyzed dynamically using multidimensional analysis tech-
niques [40]. In this way, we will identify the dimensions and the type of analysis used on
them, as indicated in Table 2.
Table 2. Dimensions, meaning and type of analysis available.
Dimension Meaning Type of Analysis Available

Sources Journals, proceedings . . . Citation, Bibliographic coupling, Co-citation
Authors Author and co-authors Co-authorship, Citation, Bibliographic coupling, Co-citation
Organizations Authors’ institutions Co-authorship, Citation, Bibliographic coupling
Countries Authors’ countries Co-authorship, Citation, Bibliographic coupling
Themes Keywords, terms, topics . . . Co-occurrence
Moreover, in this type of study, we can add temporal dimensions. Depending on the
type of analysis, we use the following bibliographic relations:
• Co-authorship analysis: The relatedness of items is determined based on their number
of co-authored documents;
• Citation analysis: The relatedness of items is determined based on the number of
times they cite each other;
• Bibliographic coupling analysis: The relatedness of items is determined based on the
number of references they share;
• Co-citation analysis: The relatedness of items is determined based on the number of
times they are cited together;
• Co-occurrence analysis: The relatedness of keywords is determined on the number
of documents in which they occur together. In this sense, the equivalence index is
usually used [41]:
c2ij
eij = (1)
ci · c j
where cij is the number of documents in which keywords i and j co-occur; ci and cj represent
the number of documents in which each appears.
𝑐2𝑖𝑗
𝑒𝑖𝑗 = (1)
𝑐𝑖 ∙ 𝑐𝑗
Mathematics 2022, 10, 1397 where cij is the number of documents in which keywords i and j co-occur; ci and cj 7repre-
of 27
sent the number of documents in which each appears.
3.5.Network
3.5. NetworkExtraction
Extractionand
andClustering
Clustering
Basedon
Based on the
the aforementioned
aforementioned relationship
relationshipmeasures
measures(Table
(Table2)2)different networks
different networks are
are constructed,
constructed, depending
depending on theon type
the type of analysis.
of analysis. After After this construction,
this construction, a process
a process of clus-
oftering
clustering or grouping
or grouping of theof the items
items that are that are considered
considered similarsimilar is usually
is usually carriedcarried out.
out. There-
Therefore, we cluster
fore, we cluster thosethose
nodes nodes
that arethatsufficiently
are sufficiently
closeclose to each
to each otherother and sufficiently
and sufficiently sepa-
separated
rated from from
the the
restrest of the
of the clusters.
clusters.
InInthis work for co-occurrence
this work for co-occurrence analysis, analysis, we we use
usethe
thesingle-link
single-linkhierarchical
hierarchicalclustering
cluster-
ing algorithm Agnes, with a network size between 3 and 12. This
algorithm Agnes, with a network size between 3 and 12. This algorithm [42] is an agglom- algorithm [42] is an
agglomerative
erative clustering clustering algorithm,
algorithm, i.e., it considers
i.e., it considers at the beginning
at the beginning that eachthat
itemeach
is a item is ain
cluster
cluster in itself,
itself, and, and,step,
in each in each step,
it tries to it tries the
group to group
nearestthe nearestorclusters
clusters or items.
items. Using the Using the
single-link
single-link
option, theoption,
distancethebetween
distancetwo between
groupstwo groups
is the is thebetween
distance distancethe between
closest the closest
individuals
individuals
in each group. in each
Othergroup.
options,Othersuchoptions, such as complete-link,
as complete-link, have been
have been discarded discarded
as they tend to
asgenerate
they tend to generate very large clusters and do not allow the identification
very large clusters and do not allow the identification of certain thematic of certain
inter-
thematic interesting areas.
esting areas.
3.6. Building of the Strategy Diagram
3.6. Building of the Strategy Diagram
The strategy diagram, Figure 3, can help to better profile the importance of each cluster
The strategy diagram, Figure 3, can help to better profile the importance of each clus-
in the co-occurrence analysis. It is based on two measures: centrality and density.
ter in the co-occurrence analysis. It is based on two measures: centrality and density.
Figure3.3.Strategy
Figure Strategydiagram.
diagram.
Theobject
The object of
of centrality will
will be
betotomeasure
measurehow
howthe networks
the networks relate to the
relate other
to the net-
other
works. This
networks. value
This valuecancanbebeunderstood
understoodasasaameasure
measure of of the importance of of an
an item
itemininthe
the
developmentofofthe
development theentire
entireresearch
researchfield
fieldanalyzed.
analyzed.ItItisisdefined
definedasasfollows
follows[41]:
[41]:
c𝑐==1010 ∑𝑒
· ∑∙ekh (2)
(2)
where ekh has been defined in Equation (1); k and h are keywords, where k is related to the
main cluster and h to other clusters.
The density can assess the internal strength of the network or of the item. This value
can be considered as a measure of the degree of the development of the item. It is defined
as follows:
∑ eij
d = 100· (3)
w
The density can assess the internal strength of the network or of the item. This value
can be considered as a measure of the degree of the development of the item. It is defined
as follows:
∑𝑒
𝑑 = 100 ∙ (3)
Mathematics 2022, 10, 1397
𝑤 8 of 27
where ekh has been defined in Equation (1); i and j are member elements of the set; and w
indicates the number of such elements within the group.
ekh has
whereOnce been
these defined in
measures Equation
have (1); i and jfor
been calculated areeach
member elements
cluster, of the
they are set; andinwa
presented
indicates
strategic the number
diagram of classifies
that such elements withininto
the themes the group.
four groups: highly developed and iso-
latedOnce theseemerging
themes; measuresorhave been calculated
declining for each
themes; basic and cluster, they themes;
transversal are presented in a
and motor
strategic
themes. diagram that classifies the themes into four groups: highly developed and isolated
themes; emerging or declining themes; basic and transversal themes; and motor themes.
3.7. Building of the Evolution and Overlapping Diagrams
3.7. Building of the Evolution and Overlapping Diagrams
We will divide the analysis into two periods: before 2016, and that year plus those
We will divide the analysis into two periods: before 2016, and that year plus those
years after it. In this way we obtain the evolution and overlapping diagram.
years after it. In this way we obtain the evolution and overlapping diagram.
For the evolution diagram we use the Inclusion Index, and for the overlapping map
For the evolution diagram we use the Inclusion Index, and for the overlapping map
we use the Jaccard Index.
we use the Jaccard Index.
In Figures 4 and 5 we can see the evolution diagram where the Inclusion Index,
In Figures 4 and 5 we can see the evolution diagram where the Inclusion Index, widely
widely used in financial analysis, is applied [43]. Figures 3 and 4 are based on the many
used in financial analysis, is applied [43]. Figures 3 and 4 are based on the many examples
examples that exist [5] about it versus others [44]. In Figure 3 there are two different evo-
that exist [5] about it versus others [44]. In Figure 3 there are two different evolution zones
lution zones separated by a line. In one is cluster A1 and cluster A2, and in the other are
separated by a line. In one is cluster A1 and cluster A2, and in the other are clusters B1, B2
clusters
and B1, solid
C2. The B2 and C2.mean
lines The solid lines
that the meanclusters
linked that theshare
linked
theclusters share the
main element. Themain ele-
dotted
ment. The dotted line means that the themes share nonmain elements. The
line means that the themes share nonmain elements. The size of the borders indicatessize of the bor-
ders
the indicatesIndex,
Inclusion the Inclusion
and the Index,
size ofand
the the size of
spheres the spheres
represents the represents
number ofthe number of
publications
publications associated
associated with the cluster.with the cluster.
Mathematics 2022, 10, x FOR PEER REVIEW 9 of 28
Figure4.4.Example
Figure Exampleof
ofevolution
evolutionareas.
areas.
Figure 4 shows the stability measures between consecutive periods (or how much
they overlap). The circles are the periods, and the numbers inside are the keywords for
each period. The horizontal arrow is the keywords shared between two consecutive peri-
ods, and the number in parentheses is the similarity index. The down arrow is the number
of outgoing keywords, and the up arrow is the number of incoming keywords in the pe-
riod.
Figure 5. Example of stability between periods.

Figure 5. Example of stability between periods.
Figure 4 shows the stability measures between consecutive periods (or how much they
3.8. Visualization
overlap). and are
The circles Interpretation
the periods,of Results
and the numbers inside are the keywords for each
period.This
Thesection is responsible
horizontal for keywords
arrow is the displayingshared
the results obtained,
between allowing theperiods,
two consecutive user to
and the number
process in parentheses
and evaluate is the similarity
them. By being index.these
able to perform The down
actionsarrow
on theisresults,
the number of
the user
outgoing
focuses on keywords,
the most and the up points
important arrow is of the
the number of incoming
dimensions keywords
and analyzes them in in the period.
detail.
In this work we used SciMAT, VOSviewer and Excel.
3.9. Overall Results of the Analysis

Taking into account all the sub-analyses carried out, an attempt is made to respond
to the objectives set out in the initial phase.
4. Application of ML in the Military Context

Mathematics 2022, 10, 1397 Figure 5. Example of stability between periods. 9 of 27
3.8. Visualization and Interpretation of Results

This sectionand
3.8. Visualization is responsible
Interpretationfor displaying the results obtained, allowing the user to
of Results
process and evaluate them. By being able to perform these actions on the results, the user
This section is responsible for displaying the results obtained, allowing the user to
focusesand
process on the most important
evaluate pointsable
them. By being of the dimensions
to perform and
these analyzes
actions them
on the in detail.
results, the user
focuses on the most important points of the dimensions and analyzes them in detail.
3.9. Overall
TakingResults of the Analysis
into account all the sub-analyses carried out, an attempt is made to respond
to the objectives
Taking set out in
into account allthe
theinitial phase. carried out, an attempt is made to respond to
sub-analyses
the objectives set out in the initial phase.
4. Application of ML in the Military Context
4. Application
Using theof ML in the Military
methodology describedContext
above, the results presented below are obtained.
First,Using thepresent
we will methodology described
the different areasabove, the we
and how results
havepresented
categorizedbelow
them,areafter
obtained.
which
First, we will present the different areas
we will offer the dimensions that emerged. and how we have categorized them, after which
we will offer the dimensions that emerged.
4.1. Theme Analysis
4.1. Theme Analysis
We present a preliminary analysis by topics in which the co-occurrence has been ap-
We
plied, present
such a preliminary
that we analysis
can see, in Figure by major
6, the topicsthemes
in which the co-occurrence
related has been
to ML. It is interesting to
applied, such that we can see, in Figure 6, the major themes related to ML. It is interesting
see the relationship that links the military environment in relation to ML with areas such
to see the relationship that links the military environment in relation to ML with areas
as AI, deep learning, algorithm and neural networks, and at the other extreme, but related
such as AI, deep learning, algorithm and neural networks, and at the other extreme, but
to the military area, with veterans, depression or the army. These areas will be discussed
related to the military area, with veterans, depression or the army. These areas will be
later.
discussed later.
Figure 6. Preliminary network by co-occurrence.

Figure 6. Preliminary network by co-occurrence.
Figure 7 establishes a clear distinction between two periods; the one prior to 2016
where publications related to ML and the military environment were limited in scope. On
the other hand, since 2016, the number of publications has increased until 2021, coinciding
with the pandemic. However, the number of citations on this topic continues to rise, which
shows that it is a topic of scientific interest. As can be seen in the section for period one,
there is hardly any distinction between the areas; however, in period two, many interesting
areas emerge.
10 of 28
Mathematics 2022, 10, 1397 10 of 27
(a)
(b)
(c)
Figure 7. Periods of analysis (a) and evolution (b) and overlay diagram (c).
Period to 2015
Focusing on the area of ML, in Figure 8, and always in relation to the military envi-
ronment, we can see a network by co-occurrence where different dimensions, such as AI,
security or intelligent systems acquire special interest in the ‘to 2015’ period.
(c)
Mathematics 2022, 10, 1397 11 of 27
Period to 2015
Period to 2015
security or
security or intelligent
intelligent systems
systems acquire
acquire special
special interest
interest in
in the
the ‘to
‘to 2015’
2015’period.
period.
Figure 8.
Figure 8. Network
Network by
by co-occurrence;
co-occurrence; ML;
ML; Period
Period to
to 2015.
2015.
Period fromfrom
Period 20162016
In the period since 2016, more categories have emerged around ML applied to the
Mathematics 2022, 10, x FOR PEER REVIEW In the period since 2016, more categories have emerged around 12 ML applied to the
of 28
military world. As we can see in Figure 9, there are different areas that can be divided into
military world. As we can see in Figure 9, there are different areas that can be divided into
five categories.
five categories.
Figure
Figure 9. Strategy
9. Strategy map;map; Categories
Categories from 2016.
from 2016.
Because of their interest, we proceed to study each of these categories in more detail:
Psychological and Behavioral Disorders

In this category we include the themes Iraq, post-traumatic stress disorder and
Mathematics 2022, 10, 1397 12 of 27
Because of their interest, we proceed to study each of these categories in more detail:
Psychological and Behavioral Disorders
In this category we include the themes Iraq, post-traumatic stress disorder and trauma.
We analyze one of the main topics, Iraq, in Figure 10, which we call Iraq and the war
in Afghanistan. Within this category, the use of data will address these disorders in relation
to the military world.
Figure10.
Figure 10. Category:
Category:Psychological andand
Psychological behavioral disorders;
behavioral FromFrom
disorders; 2016; 2016;
Theme:Theme:
Iraq andIraq
Afghan-
and
istan wars.
Afghanistan wars.
Another major
Psychological theme within
disorders: Combat thisand
category is post-traumatic
war conflicts cause stress stress
and disorder, a fre-
mental health
quent symptom
problems in veterans
in civilians, military[47]. In general,
personnel andthere are studies
veterans, whichapplying ML toinpredict
are analyzed this
different
symptomatology,
studies some ofreviews
through literature these studies
and data in analysis
Danish soldiers who participated
[45]. Behavioral disorders:inThrough
Iraq [48]
related
data to psychological
analysis, studies aredisorders.
conductedItto is determine
also easy towhich
find studies
proposedapplying analytical
solutions are thestudies
most
on prostatic stress in Afghanistan veterans related to conduct
effective for these pathologies related to the military experience [46]. disorders [49].
Finally, major
Another we discuss
themetrauma in the
within this military
category environment and
is post-traumatic how
stress there are
disorder, studies
a frequent
[50] in which
symptom ML is applied.
in veterans Some trauma
[47]. In general, data
there are analysis
studies techniques
applying ML to are more this
predict focused
symp- on
mortality [51],
tomatology, some while others
of these focus in
studies onDanish
mentalsoldiers
problems, who such as shock or
participated in stress [52].
Iraq [48] related
to psychological disorders. It is also easy to find studies applying analytical studies on
Soldier Analytics
prostatic stress in Afghanistan veterans related to conduct disorders [49].
Finally,
In this we discuss
category wetrauma
include inthe
thesoldier
military environment
theme, which inand how
itself hasthere
enoughare weight
studies to[50]
be
ina category.
which ML is applied. Some trauma data analysis techniques are more focused on
mortality
The [51],
soldierwhile
is a others focus
figure on on mental
which a largeproblems,
amount ofsuch dataasareshock or stress
analyzed, [52]. 11, and
Figure
Soldier Analytics
whose work is also facilitated by ML [53], providing an advantage, in many cases a stra-
tegicInadvantage,
this category wethe
over include
enemy, soldier
the in theme,
addition to which
creatingin itself hasenvironment
in their enough weight to be
different
aareas
category.
of study. Applications in these areas, such as resilience, prevention, diagnosis, de-
The soldier
pression, AI orisML a figure on which
can help to save;a large amount
currently, thereof are
dataalready
are analyzed,
studies Figure
focused 11,inand
this
whose worksuch
direction, is also
asfacilitated by ML [53],
clinical decision providing
support systems an advantage,
to focused,indetailed
many cases a strategicof
assessments
advantage,
suicide riskover the enemy,
in patients in addition
considered to creating
as being at highinrisk
their environment different areas of
[54].
study. Applications in these areas, such as resilience, prevention, diagnosis, depression,
Mathematics 2022, 10, 1397 13 of 27
AI or ML can help to save; currently, there are already studies focused in this direction,
such as clinical decision support systems to focused, detailed assessments of suicide risk in
patients considered as being at high risk [54].
Figure11.
Figure 11.Category:
Category:Soldier
Soldieranalytics;
analytics;From
From2016;
2016;Theme:
Theme:Soldiers.
Soldiers.
ML and Opt Opt

ML and Techniques
Techniques
In
Inthis
thiscategory
categorywe weinclude
includethethethemes
themesartificial neural
artificial network
neural network(ANN),
(ANN),deep learning,
deep learn-
boosted regression
ing, boosted tree analysis,
regression genetic
tree analysis, algorithms
genetic and adversarial
algorithms example.
and adversarial example.
The
Thearea
areaofofANN
ANNbelongs
belongsto tothe
theMLMLand
andOpt
Opttechniques’
techniques’category.
category.Through
Throughthe thearea
area
ANN
ANNand andthethecorresponding
corresponding data data processing,
processing, inin aatransversal
transversal way,
way,we wecollaborate
collaboratewith
with
different
differentareas,
areas,such
suchas assoldiers
soldiersor orpsychological
psychologicaland andbehavioral
behavioral disorders;
disorders; for
for example,
example,
studies
studies were used to highlight the variables that at the beginning of compulsorymilitary
were used to highlight the variables that at the beginning of compulsory military
service
serviceincrease
increasethethestress
stressofof
thethe
military by by
military means
meansof aof
prediction model
a prediction based
model on ANN
based [55].
on ANN
[55]. Within this category, as shown in Figure 12, we have the deep learning area, which is
a cross-cutting
Within thistheme that through
category, as shown itsin
use allows
Figure 12,the
wemilitary world
have the deeptolearning
obtain very
area,useful
which
automatic learning in their daily work, such as identifying any person anywhere
is a cross-cutting theme that through its use allows the military world to obtain very useful and
preventing crimes even before they happen [56].
automatic learning in their daily work, such as identifying any person anywhere and pre-
We see
venting within
crimes eventhe samethey
before category
happenthe[56].
boosted regression tree analysis, related to link
adaptation and K-nearest neighbors, which belongs to the scenario of highly developed
and isolated topics, being a category of little relevance.
The genetic algorithm also belongs to the same category, and we see its relationship
with optimization, reinforced learning and clustering algorithms. It can be seen how
military applications produce a large amount of data collected in the battlefield, as well
as how these data are amenable to processing by genetic algorithms, using crossover and
mutation probabilities that are automatically adjusted at each generation [45].
Adversarial example ML is a research area within this category, which focuses on the
design of strongly developed ML algorithms in adversarial environments [57].
Mathematics 2022,10,
Mathematics2022, 10,1397
x FOR PEER REVIEW 14 of
15 of27
28
Figure 12. Category: ML and Opt techniques; From 2016; Theme: Deep learning.
Figure 12. Category: ML and Opt techniques; From 2016; Theme: Deep learning.
WeRobotics
AI, see within andthe sameDevices
Smart category the boosted regression tree analysis, related to link
adaptation and K-nearest
In this category neighbors,
we include which
the themes AI,belongs to the scenario of
electroencephalogram, highly algorithm,
security, developed
and isolated topics, being a
object detection, model and big data. category of little relevance.
AIThe
is genetic
an area algorithm also belongs
of high demand in the to usetheofsame category,
data in relationand we military
to the see its relationship
world, as
with optimization, reinforced learning and clustering algorithms.
seen in this category, Figure 13. This is largely due to the increase in military It can be investment
seen how mil-
in
AIitary applications
research advances produce a large amount
[58]. It highlights the useofofdata collected
drones and theinIoT,
the tobattlefield, as well
such an extent as
that
how these
military datainare
drones someamenable
cases aretomixed
processing by genetic ones
with commercial algorithms,
[59], andusing
how crossover and
AI influences
mutation
military probabilities
strategy [60]. that are automatically adjusted at each generation [45].
Adversarial
The example ML isisarelated
electroencephalogram research to area
datawithin this category,
processing whichits
[61], although focuses on the
presence in
design of strongly developed ML algorithms
the military world is still of little significance. in adversarial environments [57].
ML algorithms collaborate with cybersecurity and, in turn, can be affected by cy-
AI, Robotics
berattacks, soand
thereSmart Devices
are more and more studies on them [62]. Technological security and
protecting data are vital in
In this category we includethe military
the environment.
themes AI, electroencephalogram, security, algo-
The need for algorithms in
rithm, object detection, model and big ML related
data. to the military world is a fact and is present in
the interpretation of the data obtained [63].
AI is an area of high demand in the use of data in relation to the military world, as
seenUnmanned aerialFigure
in this category, vehicles13.are
Thisa isreality
largely anddueinclude object detection
to the increase in militarybyinvestment
means of
algorithms;
in AI researchsome complex[58].
advances onesIt are also using
highlights the deep
use oflearning
drones and[64].the IoT, to such an extent
The model
that military drones in some cases are mixed with commercial onesarea
theme, supported primarily by design, is an emerging that
[59], andis how
beginning
AI in-
tofluences
have value with other areas,
military strategy [60]. such as trauma [51], within this category.
The military experience provides a large amount of data, big data, which need to be
analyzed by ML and is already being completed in many cases for medical purposes [65].
We can see how this area relates to data mining in the military environment in scientific
studies [66].
Military Medical Studies
In this category, we include the themes body negative pressure and training.
Mathematics 2022,10,
Mathematics2022, 10,1397
x FOR PEER REVIEW 15 of
16 of 27
28
Figure13.
Figure 13.Category:
Category:AI,
AI,robotics
roboticsand
andsmart
smartdevices;
devices;From
From2016;
2016;Theme:
Theme:Artificial
Artificialintelligence.
intelligence.
The
In electroencephalogram
relation to body negativeispressure,
related towedata processing
found evidence[61],
from although its presence in
algorithm-supported
the military
studies world data
on military
Mathematics 2022, 10, x FOR PEER REVIEW
is still of little
that focussignificance.
on military medical studies [67]. 17 of 28
ML models
The algorithms
used collaborate with cybersecurity
in military training and, processing
are based on data in turn, canandbeareaffected
becomingby
cyberattacks,common
increasingly so therein are more and
military more[68].
training studies on them [62]. Technological security and
protecting data are vital in the military environment.
4.2. Source 4.2. Source Analysis
Analysis
The need for algorithms in ML related to the military world is a fact and is present in
We can observe in Figure 14 how the number of publications related to ML applied
We can observe
the interpretation ofin
theFigure 14 how the
data obtained number of publications related to ML applied to
[63].
to the military world has clearly been increasing since 2015, although during the pandemic
the military
Unmanned worldaerial
has clearly
vehicles been
are increasing since
a reality and 2015, object
include although duringby
detection themeans
pandemic
of al-
years, it suffered a slight decline at the same time that citations continued to rise.
years,
gorithms;it suffered a slight decline
some complex ones are at also
the same
usingtime
deepthat citations
learning [64].continued to rise.
The model theme, supported primarily by design, is an emerging area that is begin-
ning to have value with other areas, such as trauma [51], within this category.
The military experience provides a large amount of data, big data, which need to be
analyzed by ML and is already being completed in many cases for medical purposes [65].
We can see how this area relates to data mining in the military environment in scientific
studies [66].
Military Medical Studies

In this category, we include the themes body negative pressure and training.
In relation to body negative pressure, we found evidence from algorithm-supported
studies on military data that focus on military medical studies [67].
The models used in military training are based on data processing and are becoming
increasingly common in military training [68].
Figure
Figure 14. Trend and 14. Trend
citation and citation structure.
structure.
The areas “Electrical-Electronic Engineering”, “Computer” and “Telecommunica-

tions” are identified as the main categories in which ML relates to the military world; see
Table 3.
Mathematics 2022, 10, 1397 16 of 27
The areas “Electrical-Electronic Engineering”, “Computer” and “Telecommunications”

are identified as the main categories in which ML relates to the military world; see Table 3.
Table 3. Main areas.
Web of Science Categories Record Count % of 525

Engineering Electrical Electronic 161 30.667
Computer Science Artificial Intelligence 119 22.667
Computer Science Information Systems 92 17.524
Computer Science Theory Methods 77 14.667
Telecommunications 77 14.667
Optics 63 12.000
Computer Science Interdisciplinary Applications 34 6.476
Automation Control Systems 24 4.571
Remote Sensing 23 4.381
Instruments Instrumentation 21 4.000
Imaging Science Photographic Technology 20 3.810
Engineering Multidisciplinary 19 3.619
Psychiatry 19 3.619
Computer Science Software Engineering 18 3.429
Physics Applied 17 3.238
Chemistry Analytical 16 3.048
Computer Science Cybernetics 16 3.048
Computer Science Hardware Architecture 16 3.048
Engineering Aerospace 15 2.857
Operations Research Management Science 13 2.476
Robotics 13 2.476
Environmental Sciences 12 2.286
Materials Science Multidisciplinary 11 2.095
Neurosciences 11 2.095
Surgery 11 2.095
4.3. Country Analysis

Once the number of citations and documents published on ML oriented to the military
world by country is normalized, we proceed to their comparison, as we can see in Figure 15.
We can observe that, in some cases, the number of publications and citations do correspond,
as in the case of China, but in other cases, such as England, Israel and Iran, the number of
citations is much higher than the number of publications, which means that the quality of
these publications is very high. The opposite is the case in India, South Korea and Canada,
where the number of citations is much lower than the number of publications, meaning
low quality.
In Figure 16, we can see in general terms the relationship between different countries
in terms of joint participation in the writing of scientific papers on ML applied to the
world of work, and therefore the relationship of the co-authorship of researchers from
different countries in the writing of these papers. The size of each circuit is proportional
to the number of papers, and subsequently we will identify the main countries in the
production of scientific material in this topic during the period of highest productivity:
2016–2019. The USA co-author network has been changing in recent years, starting in 2016
with a collaboration network mostly with France and the Netherlands, followed in 2017
and 2018 mainly with England, Australia, Greece and continuing with the Netherlands,
until finally in 2019 the main relationship is with China, India and Taiwan. We can see
how England centered in 2016 its network of co-authors in France and the Netherlands,
collaborating in 2017 and 2018 with the USA, Australia, Israel, Iran and Turkey to finally
establish stronger co-authorship with China, India or Saudi Arabia, starting in 2019. China
shares co-authorship in ML scientific publications in the military world with Norway in
2016, evolving with the USA, Australia and England during 2017 and 2018 until 2019, when
the relationship with India, Canada, Saudi Arabia or Turkey is strengthened.
tary world by country is normalized, we proceed to their comparison, as we can see in
Figure 15. We can observe that, in some cases, the number of publications and citations
do correspond, as in the case of China, but in other cases, such as England, Israel and Iran,
the number of citations is much higher than the number of publications, which means that
the quality of these publications is very high. The opposite is the case in India, South Korea
Mathematics 2022, 10, 1397 17 of 27
and Canada, where the number of citations is much lower than the number of publica-
tions, meaning low quality.

Figure 15. Country
Country analysis;
analysis; Most important by documents and citations.
In Figure 16, we can see in general terms the relationship between different countries
in terms of joint participation in the writing of scientific papers on ML applied to the world
of work, and therefore the relationship of the co-authorship of researchers from different
countries in the writing of these papers. The size of each circuit is proportional to the
number of papers, and subsequently we will identify the main countries in the production
of scientific material in this topic during the period of highest productivity: 2016–2019.
The USA co-author network has been changing in recent years, starting in 2016 with a
collaboration network mostly with France and the Netherlands, followed in 2017 and 2018
mainly with England, Australia, Greece and continuing with the Netherlands, until finally
in 2019 the main relationship is with China, India and Taiwan. We can see how England
centered in 2016 its network of co-authors in France and the Netherlands, collaborating in
2017 and 2018 with the USA, Australia, Israel, Iran and Turkey to finally establish stronger
co-authorship with China, India or Saudi Arabia, starting in 2019. China shares co-author-
ship in ML scientific publications in the military world with Norway in 2016, evolving
with the USA, Australia and England during 2017 and 2018 until 2019, when the relation-
ship with India, Canada, Saudi Arabia or Turkey is strengthened.
Figure 16.
Figure 16. Network
Network by
by co-authorship.
co-authorship.
4.4.
4.4. Organization
Organization Analysis
Analysis
In
In Figure 17, we
Figure 17, we normalize
normalize by by organization
organization thethe number
number ofof scientific
scientific papers
papers onon ML
ML
applied to the military world in relation to different organizations.
applied to the military world in relation to different organizations.
It
It is
is striking
striking to
to see
see how
how few
few citations
citations the
the US
US army
army organization
organization hashas in
in relation
relation to
to the
the
number
number of of publications
publications itit generates;
generates; the
the same
same happens
happens with
with the
the Korea
Korea Advanced
Advanced Institute
Institute
of
of Science
Science and and Technology,
Technology, the the Pennsylvania
Pennsylvania State
State University,
University, Sejong
Sejong University,
University, Florida
Florida
State University, Georgia Institute of Technology and Korea
State University, Georgia Institute of Technology and Korea University.University.
On the other hand, we can see the high number of citations of organizations that have
published to a lesser extent, which means that these publications are of high quality, such
as the Uniformed Services University of the Health Sciences, Harvard Medical School, the
University of Pittsburgh, Boston University, VA San Diego Healthcare System, the Uni-
versity of California San Diego and the University of Michigan.
of Science and Technology, the Pennsylvania State University, Sejong University, Florida
State University, Georgia Institute of Technology and Korea University.
published to a lesser extent, which means that these publications are of high quality, such
as the Uniformed Services University of the Health Sciences, Harvard Medical School, the
Mathematics 2022, 10, 1397 18 of 27
University of Pittsburgh, Boston University, VA San Diego Healthcare System, the Uni-
versity of California San Diego and the University of Michigan.
Figure
Figure 17.
17. Org
Org analysis;
analysis; Most
Most important by documents
important by documents and
and citations.
citations.
published
Mathematics 2022, 10, x FOR PEER REVIEW
to a lesser extent, which means that these publications are of high quality, such
20 of 28
as the Uniformed Services University of the Health Sciences, Harvard Medical School,
the University of Pittsburgh, Boston University, VA San Diego Healthcare System, the
University of California San Diego and the University of Michigan.
Wepresent
We presentininFigure
Figure1818the
the co-authorship
co-authorship network
network between
between the the different
different organiza-
organizations
tions in the most significant period, between 2018 and 2019. First, we present
in the most significant period, between 2018 and 2019. First, we present where the most where the
most significant organizations are: The US Army establishes a relationship
significant organizations are: The US Army establishes a relationship in 2018 with the in 2018 with
the Uniformed
Uniformed Services
Services University
University of of
thethe Health
Health Sciences,the
Sciences, theUniversity
Universityof ofCalifornia
California San
San
Diego and the University of Virginia to to generate
generate scientific
scientific articles on ML applied
applied to the
the
military world. This relationship continues with the University of Colorado and Harvard
University, ending
University, ending with
with Emory
Emory University
University inin 2019;
2019; the
the Uniformed
Uniformed Services
Services University
University of
the Health
the Health Sciences
Sciences begins
begins its co-authorship
co-authorship in 2018 with several organizations, but, before
long, in 2019 it focuses its relationships primarily on the Johns Hopkins University,
long, in this
University, in
being their
case being their only
only relationship
relationship byby mid-2019
mid-2019 (Figure
(Figure 19).
19).
Figure 18. Network by co-authorship.

Mathematics 2022, 10, 1397 19 of 27
Figure
Figure19.
19.Network
Networkby
bycitation.
citation.
Bibliographic
Bibliographiccollaboration
collaborationbetween
betweendifferent
differentorganizations
organizationswas
wasvery
verystrong
strongatat the
the
atics 2022, 10, x FOR PEER REVIEW beginning of 2018, and a decrease is seen at the beginning of 2019, as can
21 of be
28 seen
beginning of 2018, and a decrease is seen at the beginning of 2019, as can be seen in in Figure
20.
Figure 20.
Figure 20. Network by bibliographic

Figure 20. Network bycoupling.
bibliographic coupling.
4.5. Author Analysis

Inthe
In an analysis by an authors,
analysis by
wethe authors,the
recognize wemost
recognize theauthors
relevant most relevant
by theirauthors
numberby their number
of publications, citations and year, highlighting, in 2018, Ben-David
of publications, citations and year, highlighting, in 2018, Ben-David for the highest fornum-
the highest number
of citations, as we can see in Figure 21. Arie Ben-David works at
ber of citations, as we can see in Figure 21. Arie Ben-David works at the Holon Institutethe Holon Institute of
Technology, Holon, Israel, and has more than 24 publications and
of Technology, Holon, Israel, and has more than 24 publications and more than 850 cita- more than 850 citations
to their
tions to their name. name.
In an analysis by the authors, we recognize the most relevant authors by their number
of publications, citations and year, highlighting, in 2018, Ben-David for the highest num-
ber of citations, as we can see in Figure 21. Arie Ben-David works at the Holon Institute
Mathematics 2022, 10, 1397 of Technology, Holon, Israel, and has more than 24 publications and more than 85020cita-of 27
tions to their name.
Figure 21. Network by citations.

Figure 21. Network by citations.

The exponential evolution of the number of publications possibly slowed down from
2020 due to the effects of COVID-19. The most interesting analysis corresponds to the period
since 2016. There is a predominance of journals on engineering, electricals, electronics and
computing sciences [69].
The USA, China, the UK and South Korea lead in publications and citations. There
is a predominance of US universities and the US Army in publications and citations. The
main thematic categories have been identified and their importance and perspectives have
been characterized:
• Psychological and behavioral disorders;
• Soldier analytics;
• ML and Opt techniques;
• AI, robotics and smart devices;
• Military medical studies.
5. Discussion
Based on the bibliometric analysis performed, in this section, we redefine the concep-
tual architecture presented in Section 2.2. by adapting it to the military context, as shown
in Figure 22.
The components of each layer are explained below:
– Data management solutions for military analytics (DMSMA). In general terms, it can
be said that the data that predominate in the military context are more complex to
process than those in a conventional organization. In the study we have carried out,
the main sources found are biometric data, e.g., from electroencephalograms [70]; data
from military experts and commanders (e.g., field assessments and weaponry needed
in that particular field), which should be used in conjunction with the other data to
build decision systems [35]; the Internet of things, smart and connected devices widely
used by the military, generating large volumes of information over time [71], e.g., those
provided by radars [72]; the Internet of battlefield things, which connects soldiers
with smart technology in weapons and other objects to give troops “extra sensory
powers” [73]; military personnel data, e.g., those obtained in screening interviews [74];
Mathematics 2022, 10, 1397 21 of 27

military veterans’ data obtained, e.g., from their administrative personal files [75];
and other data, both internal and open in any format (including videos, images and
speeches), e.g., those coming from unmanned aerial vehicles [76], videos for facial
4.6. Overall Results[77,78],
recognition of the Analysis
data from the Internet, etc.;
– The Insight generation for
exponential evolution military.
of the The
numberpredominant ML algorithms
of publications in this layer
possibly slowed downare the
from
so-called deep ML algorithms, since these algorithms give better
2020 due to the effects of COVID-19. The most interesting analysis corresponds to the pe- results with a large
volume
riod since 2016.ofThere
data and/or unstructured
is a predominance of data [69].on
journals These algorithms
engineering, are based
electricals, on neural
electronics
networks [79]. If
and computing sciences [69]. the number of layers of such a network is high, they are called deep
learning, e.g., these algorithms are applied to military subjects
The USA, China, the UK and South Korea lead in publications and citations. There is in [72,80,81]. Deep
adversarial of
a predominance algorithms are oftenand
US universities used to US
the attack
Armyother
inML models and
publications cause
and their failure,
citations. The
e.g., [82]. Another deep variant is the long short-term memory
main thematic categories have been identified and their importance and perspectives (LSTM) algorithms,
have
which are specialized in the treatment of large-scale time series, e.g., [83]. Another type
been characterized:
of deep network is the convolutional neural networks (CNNs) that are often used for
● Psychological
object detectionand[82].
behavioral disorders;
In contrast to deep algorithms, there are shallow algorithms [83];
● Soldier analytics;
application examples can be found in [84]. We can also find examples of clustering [85],
● MLtheand Opt techniques;
detection of outliers [86], optimization [87], reinforcement learning [88], etc.;
●– AI, robotics and smart devices;
Military application. Several of these applications have already been discussed in
● Military
Section medical studies. is a list of the most important ones: cybersecurity and threat
4.1. The following
intelligence [88]; image, speech and pattern recognition [89]; the mental health of sol-
5. Discussion
diers and veterans [74]; military ethics [90]; military personnel behavior analytics and
Based on the bibliometric
management analysis
[91,92]; military performed,
robotics and smartin this section,
devices [63];we redefine
and the concep-
the physical health
of soldiers and
tual architecture veterans,
presented etc. It is2.2.
in Section remarkable
by adapting thatitseveral of these context,
to the military ML applications
as shownin
the 22.
in Figure military field have not been identified in previous works [36].
Figure 22. Data-driven architecture for military organizations.

Figure 22. Data-driven architecture for military organizations.
The
Thecomponents
componentsofofeach
eachlayer
layerare
areexplained
explainedbelow:
below:
- Data management solutions for military analytics (DMSMA). In general terms, it can
the main sources found are biometric data, e.g., from electroencephalograms [70];
Mathematics 2022, 10, 1397 22 of 27
– Data management solutions for military analytics (DMSMA). In general terms, it can
the main sources found are biometric data, e.g., from electroencephalograms [70]; data
from military experts and commanders (e.g., field assessments and weaponry needed
in that particular field), which should be used in conjunction with the other data to
build decision systems [35]; the Internet of things, smart and connected devices widely
used by the military, generating large volumes of information over time [71], e.g., those
provided by radars [72]; the Internet of battlefield things, which connects soldiers
with smart technology in weapons and other objects to give troops “extra sensory
powers” [73]; military personnel data, e.g., those obtained in screening interviews [74];
military veterans’ data obtained, e.g., from their administrative personal files [75];
and other data, both internal and open in any format (including videos, images and
speeches), e.g., those coming from unmanned aerial vehicles [76], videos for facial
recognition [77,78], data from the Internet, etc.;
– Insight generation for military. The predominant ML algorithms in this layer are the
so-called deep ML algorithms, since these algorithms give better results with a large
volume of data and/or unstructured data [69]. These algorithms are based on neural
networks [79]. If the number of layers of such a network is high, they are called deep
learning, e.g., these algorithms are applied to military subjects in [72,80,81]. Deep
adversarial algorithms are often used to attack other ML models and cause their failure,
e.g., [82]. Another deep variant is the long short-term memory (LSTM) algorithms,
which are specialized in the treatment of large-scale time series, e.g., [83]. Another type
of deep network is the convolutional neural networks (CNNs) that are often used for
object detection [82]. In contrast to deep algorithms, there are shallow algorithms [83];
application examples can be found in [84]. We can also find examples of clustering [85],
the detection of outliers [86], optimization [87], reinforcement learning [88], etc.;
– Military application. Several of these applications have already been discussed in
Section 4.1. The following is a list of the most important ones: cybersecurity and threat
intelligence [88]; image, speech and pattern recognition [89]; the mental health of sol-
diers and veterans [74]; military ethics [90]; military personnel behavior analytics and
management [91,92]; military robotics and smart devices [63]; and the physical health
of soldiers and veterans, etc. It is remarkable that several of these ML applications in
the military field have not been identified in previous works [36].
6. Conclusions and Future Work

The objective of this work has been fulfilled: we have carried out a research work
presenting initially an ML architecture model applied to a nonmilitary organization, after
which a bibliometric study on the use of ML applied to military organizations is presented
and, finally, we have applied this study to the original model to obtain an architecture
model to apply ML to a military organization. All this has been executed while taking into
account that previously there was a lack of scientific information on this subject.
A clear map of the present, past and future of research has been provided.
It has shown a real application of ML and the growing real interest in applying it, in
this case to the military field, observing how it is increasingly used to analyze data for
automatic decision making.
The following aspects are highlighted:
• The wars in Afghanistan and Iraq are coming under intense scrutiny for their mental
effects on military personnel;
• Soldier analytics could become an area in its own right, as it has a lot of specificity
compared to today’s people analytics;
• The area of deep learning is growing in military applications;
• There are interesting emerging topics related to AI, such as intrusion detection, brain
interfaces, self-driving vehicles, false information processing, cybersecurity, etc.;
Mathematics 2022, 10, 1397 23 of 27
• Underlying this is an intrinsic importance in medical research that is likely to have

fewer strategic constraints from governments.
We encountered the following limitations: we have not had access to specialized
military libraries, and bias is assumed in several publications due to the subject matter.
As lines of future work, the following is highlighted as having a longer development
path. The usefulness of ML for the management of military personnel in the style of other
already-consolidated areas, such as people analytics [93], has been proven. However,
this new and specific area, which we could call soldier analytics, is not yet defined in
the literature.
Author Contributions: Conceptualization, J.J.G. and R.A.C.; methodology, J.J.G. and R.A.C.; software,
J.J.G. and R.A.C.; validation, J.J.G., R.A.C. and A.L.; formal analysis, J.J.G. and R.A.C.; investigation,
J.J.G.; resources, J.J.G., R.A.C. and A.L.; data curation, J.J.G. and R.A.C.; writing—original draft prepa-
ration, J.J.G.; writing—review and editing, J.J.G., R.A.C. and A.L.; visualization, J.J.G.; supervision,
A.L. and R.A.C.; project administration, J.J.G. and R.A.C. All authors have read and agreed to the
published version of the manuscript.
Funding: This research has been partially supported support from the FEDER funds provided by the
National Spanish project PGC2018-096509-B-I00.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Acknowledgments: The authors would like to acknowledge the financial support from the FEDER
funds provided by the National Spanish project PGC2018-096509-B-I00 and the DATA SCIENCE AND
SOFT COMPUTING FOR SOCIAL ANALYTICS AND DECISION AID Research Group. LaTorre
thanks the Spanish Ministry of Science (grant PID2020-113013RB-C22). Carrasco thanks the Spanish
Ministry of Science (grant PID2019-103880RB-I00).
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Zhang, Y.; Dai, Z.; Zhang, L.; Wang, Z.; Chen, L.; Zhou, Y. Application of Artificial Intelligence in Military: From Projects View.
In Proceedings of the 2020 6th International Conference on Big Data and Information Analytics (BigDIA), Shenzhen, China,
4–6 December 2020; pp. 113–116. [CrossRef]
2. Pan, J.; Yi, K.; Fang, H.; Zhao, Y.; Du, K.; Liu, W. Analysis of Military Information Construction under the Background of Artificial
Intelligence. E3S Web Conf. 2021, 235, 03046. [CrossRef]
3. Cho, S.; Shin, W.; Kim, N.; Jeong, J.; In, H.P. Priority Determination to Apply Artificial Intelligence Technology in Military
Intelligence Areas. Electronics 2020, 9, 2187. [CrossRef]
4. Ajakwe, S.O.; Nwakanma, C.I.; Lee, J.M.; Kim, D.S. Machine Learning Algorithm for Intelligent Prediction for Military Logistics
and Planning. In Proceedings of the 2020 International Conference on Information and Communication Technology Convergence
(ICTC), Jeju, Korea, 21–23 October 2020; pp. 417–419. [CrossRef]
5. Cobo, M.J.; López-Herrera, A.G.; Herrera-Viedma, E.; Herrera, F. SciMAT: A new science mapping analysis software tool. J. Am.
Soc. Inf. Sci. Technol. 2012, 63, 1609–1630. [CrossRef]
6. Azar, A.T.; Koubaa, A.; Mohamed, N.A.; Ibrahim, H.A.; Ibrahim, Z.F.; Kazim, M.; Ammar, A.; Benjdira, B.; Khamis, A.M.;
Hameed, I.A.; et al. Drone Deep Reinforcement Learning: A Review. Electronics 2021, 10, 999. [CrossRef]
7. Pundir, M.; Sandhu, J.K. A Systematic Review of Quality of Service in Wireless Sensor Networks using Machine Learning: Recent
Trend and Future Vision. J. Netw. Comput. Appl. 2021, 188, 103084. [CrossRef]
8. Liu, H.; Meng, G.; Deng, Z.; Li, M.; Chang, J.; Dai, T.; Fang, X. Progress in Research on VOC Molecule Recognition by
Semiconductor Sensors. Acta Physico-Chim. Sin. 2020, 38, 2008018. [CrossRef]
9. Mellit, A.; Kalogirou, S.A. MPPT-based artificial intelligence techniques for photovoltaic systems and its implementation into
field programmable gate array chips: Review of current status and future perspectives. Energy 2014, 70, 1–21. [CrossRef]
10. Wilson, A.N.; Kumar, A.; Jha, A.; Cenkeramaddi, L.R. Embedded Sensors, Communication Technologies, Computing Platforms
and Machine Learning for UAVs: A Review. IEEE Sens. J. 2021, 22, 1807–1826. [CrossRef]
11. Al-Kaff, A.; Martín, D.; García, F.; de la Escalera, A.; Armingol, J.M. Survey of computer vision algorithms and applications for
unmanned aerial vehicles. Expert Syst. Appl. 2018, 92, 447–463. [CrossRef]
12. Li, Y.; Zhang, H.; Xue, X.; Jiang, Y.; Shen, Q. Deep learning for remote sensing image classification: A survey. WIREs Data Min.
Knowl. Discov. 2018, 8, 1264. [CrossRef]
Mathematics 2022, 10, 1397 24 of 27
13. Zhang, Y.; Xin, J. Survivable Deployments of Optical Sensor Networks against Multiple Failures and Disasters: A Survey. Sensors
2019, 19, 4790. [CrossRef] [PubMed]
14. Convertino, V.A.; Schauer, S.G.; Weitzel, E.K.; Cardin, S.; Stackle, M.E.; Talley, M.J.; Sawka, M.N.; Inan, O.T. Wearable Sensors
Incorporating Compensatory Reserve Measurement for Advancing Physiological Monitoring in Critically Injured Trauma Patients.
Sensors 2020, 20, 6413. [CrossRef] [PubMed]
15. Martins, O.; Adekunle, A.A.; Adejuyigbe, S.B.; Adeyemi, H.O. Quantitative Performance Review of Wheeled Mobile Robot Path
Planning Algorithms. GAZI Univ. J. Sci. 2021, 34, 765–784. [CrossRef]
16. Banerjee, A.; Maity, S.; Mastrangelo, C.H. Nanostructures for Biosensing, with a Brief Overview on Cancer Detection, IoT, and the
Role of Machine Learning in Smart Biosensors. Sensors 2021, 21, 1253. [CrossRef]
17. Tian, Y.; Lan, L.; Guo, H. A review on the wavelet methods for sonar image segmentation. Int. J. Adv. Robot. Syst. 2020, 17,
1729881420936091. [CrossRef]
18. Aziz, L.; Salam, M.S.B.H.; Sheikh, U.U.; Ayub, S. Exploring Deep Learning-Based Architecture, Strategies, Applications and
Current Trends in Generic Object Detection: A Comprehensive Review. IEEE Access 2020, 8, 170461–170495. [CrossRef]
19. Nampoothiri, M.G.; Vinayakumar, B.; Sunny, Y.; Antony, R. Recent developments in terrain identification, classification, parameter
estimation for the navigation of autonomous robots. SN Appl. Sci. 2021, 3, 480. [CrossRef]
20. Zhang, H.; Xin, B.; Dou, L.-H.; Chen, J.; Hirota, K. A review of cooperative path planning of an unmanned aerial vehicle group.
Front. Inf. Technol. Electron. Eng. 2020, 21, 1671–1694. [CrossRef]
21. Hu, W.-D.; Li, Q.; Chen, X.-S.; Lu, W. Recent progress on advanced infrared photodetectors. Acta Phys. Sin. 2019, 68, 20190281.
[CrossRef]
22. Degenstein, L.M.; Sameoto, D.; Hogan, J.D.; Asad, A.; Dolez, P.I. Smart Textiles for Visible and IR Camouflage Application:
State-of-the-Art and Microfabrication Path Forward. Micromachines 2021, 12, 773. [CrossRef]
23. de Neufville, R.; Baum, S.D. Collective action on artificial intelligence: A primer and review. Technol. Soc. 2021, 66, 101649.
[CrossRef]
24. Amalina, F.; Hashem, I.A.T.; Azizul, Z.H.; Fong, A.T.; Firdaus, A.; Imran, M.; Anuar, N.B. Blending Big Data Analytics: Review on
Challenges and a Recent Study. IEEE Access 2019, 8, 3629–3645. [CrossRef]
25. Evangelista, J.R.G.; Sassi, R.J.; Romero, M.; Napolitano, D. Systematic Literature Review to Investigate the Application of Open
Source Intelligence (OSINT) with Artificial Intelligence. J. Appl. Secur. Res. 2021, 16, 345–369. [CrossRef]
26. De las Heras-Rosas, C.D.L.; Herrera, J. Innovation and Competitive Intelligence in Business. A Bibliometric Analysis. Int. J.
Financ. Stud. 2021, 9, 31. [CrossRef]
27. Dodge, J.; Khanna, R.; Irvine, J.; Lam, K.-H.; Mai, T.; Lin, Z.; Kiddle, N.; Newman, E.; Anderson, A.; Raja, S.; et al. After-Action
Review for AI (AAR/AI). ACM Trans. Interact. Intell. Syst. 2021, 11, 1–35. [CrossRef]
28. Samarasinghe, D.; Barlow, M.; Lakshika, E.; Lynar, T.; Moustafa, N.; Townsend, T.; Turnbull, B. A Data Driven Review of Board
Game Design and Interactions of Their Mechanics. IEEE Access 2021, 9, 114051–114069. [CrossRef]
29. Khan, M.J.; Khan, H.S.; Yousaf, A.; Khurshid, K.; Abbas, A. Modern Trends in Hyperspectral Image Analysis: A Review. IEEE
Access 2018, 6, 14118–14129. [CrossRef]
30. Small, S.G.; Medsker, L. Review of information extraction technologies and applications. Neural Comput. Appl. 2014, 25, 533–548.
[CrossRef]
31. López-Robles, J.R.; Otegi-Olaso, J.R.; Gómez, I.P.; Cobo, M.J. 30 years of intelligence models in management and business: A
bibliometric review. Int. J. Inf. Manag. 2019, 48, 22–38. [CrossRef]
32. Miller, D.J.; Xiang, Z.; Kesidis, G. Adversarial Learning Targeting Deep Neural Network Classification: A Comprehensive Review
of Defenses Against Attacks. Proc. IEEE 2020, 108, 402–433. [CrossRef]
33. Gomes, G.F.; Mendéz, Y.A.D.; Alexandrino, P.D.S.L.; da Cunha, S.S., Jr.; Ancelotti, A.C., Jr. The use of intelligent computational
tools for damage detection and identification with an emphasis on composites–A review. Compos. Struct. 2018, 196, 44–54.
[CrossRef]
34. Rempel, M.; Cai, J. A review of approximate dynamic programming applications within military operations research. Oper. Res.
Perspect. 2021, 8, 100204. [CrossRef]
35. Wang, W.; Liu, H.; Lin, W.; Chen, Y.; Yang, J.-A. Investigation on Works and Military Applications of Artificial Intelligence. IEEE
Access 2020, 8, 131614–131625. [CrossRef]
36. Svenmarck, P.; Luotsinen, L.; Nilsson, M.; Schubert, J. Possibilities and challenges for artificial intelligence in military applications.
In Proceedings of the NATO Big Data and Artificial Intelligence for Military Decision Making Specialists’ Meeting, Bordeaux,
France, 30 May–1 June 2018; pp. 1–16.
37. Moreno, C.; Carrasco, R.A.; Herrera-Viedma, E. Data and Artificial Intelligence Strategy: A Conceptual Enterprise Big Data Cloud
Architecture to Enable Market-Oriented Organisations. Int. J. Interact. Multimed. Artif. Intell. 2019, 5, 6. [CrossRef]
38. Xu, Y.; Zhou, Y.; Sekula, P.; Ding, L. Machine learning in construction: From shallow to deep learning. Dev. Built Environ. 2021, 6,
100045. [CrossRef]
39. Sarker, I.H. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN Comput. Sci. 2021, 2, 160.
[CrossRef] [PubMed]
40. Cid-López, A.; Hornos, M.J.; Carrasco, R.A.; Herrera-Viedma, E. A hybrid model for decision-making in the Information and
Communications Technology sector. Technol. Econ. Dev. Econ. 2015, 21, 720–737. [CrossRef]
Mathematics 2022, 10, 1397 25 of 27
41. Cobo, M.J.; Jürgens, B.; Herrero-Solana, V.; Martínez, M.A.; Herrera-Viedma, E. Industry 4.0: A perspective based on bibliometric
analysis. Procedia Comput. Sci. 2018, 139, 364–371. [CrossRef]
42. Huang, W.-Q.; Ding, C.; Wang, S.-Y.; Hu, S. An Efficient Clustering Mining Algorithm for Indoor Moving Target Trajectory Based
on the Improved AGNES. IEEE 2015, 1, 1318–1323. [CrossRef]
43. Cámara, N.; Tuesta, D. Measuring financial inclusion: A muldimensional index. BBVA Res. Pap. 2014, 14, 26. [CrossRef]
44. Verma, V.; Aggarwal, R.K. A comparative analysis of similarity measures akin to the Jaccard index in collaborative recommenda-
tions: Empirical and theoretical perspective. Soc. Netw. Anal. Min. 2020, 10, 1–16. [CrossRef]
45. Kehle-Forbes, S.M.; Campbell, E.H.; Taylor, B.C.; Scholten, J.; Sayer, N. Does Co-Occurring Traumatic Brain Injury Affect VHA
Outpatient Health Service Utilization and Associated Costs Among Veterans with Posttraumatic Stress Disorder? An Examination
Based on VHA Administrative Data. J. Head Trauma Rehabil. 2017, 32, E16–E23. [CrossRef] [PubMed]
46. Adamowski, T.; Kiejna, A. Mental Disorders Between Military Conflict’s Veterans. Adv. Clin. Exp. Med. 2006, 15, 355–360.
47. Kang, B.; Karel, M.J.; Corazzini, K.N.; McConnell, E.S. A mixed methods study on the manifestations of behavioural symptoms of
dementia among veterans with and without posttraumatic stress disorder. J. Adv. Nurs. 2021, 77, 3176–3188. [CrossRef] [PubMed]
48. Goderez, B.I. The survivor syndrome. Massive psychic trauma and posttraumatic stress disorder. Bull. Menn. Clin. 1987, 51, 96.
49. Karstoft, K.-I.; Statnikov, A.; Andersen, S.B.; Madsen, T.; Galatzer-Levy, I.R. Early identification of posttraumatic stress following
military deployment: Application of machine learning methods to a prospective study of Danish soldiers. J. Affect. Disord. 2015,
184, 170–175. [CrossRef]
50. Pietrzak, R.H.; Tsai, J.; Harpaz-Rotem, I.; Whealin, J.M.; Southwick, S.M. Support for a novel five-factor model of posttraumatic
stress symptoms in three independent samples of Iraq/Afghanistan veterans: A confirmatory factor analytic study. J. Psychiatr.
Res. 2012, 46, 317–322. [CrossRef]
51. Lammers, D.; Marenco, C.; Morte, K.; Conner, J.; Williams, J.; Bax, T.; Martin, M.; Eckert, M.; Bingham, J. Machine Learning for
Military Trauma: Novel Massive Transfusion Predictive Models in Combat Zones. J. Surg. Res. 2022, 270, 369–375. [CrossRef]
52. Bickel, K.E.; Kennedy, R.; Levy, C.; Burgio, K.L.; Bailey, F.A. The Relationship of Post-traumatic Stress Disorder to End-of-life Care
Received by Dying Veterans: A Secondary Data Analysis. J. Gen. Intern. Med. 2020, 35, 505–513. [CrossRef]
53. Lance, B.J.; Larkin, G.B.; Touryan, J.O.; Rexwinkle, J.T.; Gutstein, S.M.; Gordon, S.M.; Toulson, O.; Choi, J.; Mahdi, A.;
Hung, C.P.; et al. Minimizing data requirements for soldier-interactive AI/ML applications through opportunistic sensing.
Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications II. Int. Soc. Opt. Photonics 2020, 11413,
1141306. [CrossRef]
54. Zuromski, K.L.; Bernecker, S.L.; Gutierrez, P.M.; Joiner, T.E.; King, A.J.; Liu, H.; Naifeh, J.A.; Nock, M.K.; Sampson, N.A.;
Zaslavsky, A.M.; et al. Assessment of a Risk Index for Suicide Attempts Among US Army Soldiers with Suicide Ideation: Analysis
of Data From the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). JAMA Netw. Open 2019, 2,
e190766. [CrossRef] [PubMed]
55. Bekesiene, S.; Smaliukiene, R.; Vaicaitiene, R. Using Artificial Neural Networks in Predicting the Level of Stress among Military
Conscripts. Mathmatics 2021, 9, 626. [CrossRef]
56. Aouto, A.; Huynh-The, T.; Lee, J.-M.; Kim, D.-S. Pose-Based Identification Using Deep Learning for Military Surveillance Systems.
In Proceedings of the 2019 International Conference on Information and Communication Technology Convergence (ICTC), Jeju
Island, Korea, 16–18 October 2019; pp. 626–629.
57. Duddu, V. A survey of adversarial machine learning in cyber warfare. Def. Sci. J. 2018, 68, 356. [CrossRef]
58. Fazekas, F. AI and Military Operations’ Planning. In Artificial Intelligence and Its Contexts; Springer: Berlin/Heidelberg, Germany,
2021; pp. 79–91.
59. Restituyo, R.; Hayajneh, T. Vulnerabilities and Attacks Analysis for Military and Commercial IoT Drones. In Proceedings of the
2018 9th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), New York, NY,
USA, 8–10 November 2018; pp. 26–32.
60. Ploumis, M. AI weapon systems in future war operations; strategy, operations and tactics. Comp. Strat. 2022, 41, 1–18. [CrossRef]
61. Srinath, R.; Gayathri, R. Detection and classification of electroencephalogram signals for epilepsy disease using machine learning
methods. Int. J. Imaging Syst. Technol. 2021, 31, 729–740. [CrossRef]
62. Chukhnov, A.P.; Ivanov, Y.S. Algorithms for Detecting and Preventing Attacks on Machine Learning Models in Cyber-Security
Problems. J. Phys. Conf. Ser. 2021, 2096, 012099. [CrossRef]
63. Kang, J.J. A Military Human Performance Management System Design using Machine Learning Algorithms. In Proceedings
of the 2021 31st International Telecommunication Networks and Applications Conference (ITNAC), Sydney, Australia, 24–26
November 2021; pp. 13–18. [CrossRef]
64. Escobar-Flores, J.G.; Sandoval, S.; Gámiz-Romero, E. Unmanned aerial vehicle images in the machine learning for agave detection.
Environ. Sci. Pollut. Res. 2022, 1–12. [CrossRef]
65. Berry, D. Application of big data in healthcare: Examination of the military experience. Health Technol. 2021, 11, 251–256.
[CrossRef]
66. Ren, L.; Chen, Y. Research on the Application of Data Mining Technology in Military Audit. In Proceedings of the 2021
International Conference on Education, Information Management and Service Science (EIMSS), Xi’an, China, 16–18 July 2021;
pp. 277–283.
Mathematics 2022, 10, 1397 26 of 27
67. Walker, P.B.; Norris, J.N.; Tschiffely, A.E.; Mehalick, M.L.; Cunningham, C.A.; Davidson, I.N.; Tshiffiley, A.E. Applications of
Transductive Spectral Clustering Methods in a Military Medical Concussion Database. IEEE/ACM Trans. Comput. Biol. Bioinform.
2016, 14, 534–544. [CrossRef]
68. Zheng, Y.; Li, W.; Dong, X.; Zhou, M.; Xu, L. Application Research on Big Data of Military Training in Military Academy Teaching.
In Proceedings of the 2020 International Conference on Modern Education and Information Management (ICMEIM), Dalian,
China, 25–27 September 2020; pp. 357–361.
69. Talbot, L.A.; Haffner, W.H.J.; Rice, C.L.; Rothwell, S.W. COVID-19 Pandemic: A Collection of Relevant Publications from Military
Medicine. Mil. Med. 2020, 185, e1364–e1365. [CrossRef]
70. Zhang, Y.; Xu, P.; Li, P.; Duan, K.; Wen, Y.; Yang, Q.; Zhang, T.; Yao, D. Noise-assisted multivariate empirical mode decomposition
for multichannel EMG signals. Biomed. Eng. Online 2017, 16, 107. [CrossRef] [PubMed]
71. Hanson, P.; Truax, L.; Saranchak, D.D. IOT honeynet for military deception and indications and warnings. In Autonomous
Systems: Sensors, Vehicles, Security, and the Internet of Everything. Int. Soc. Opt. Photonics 2018, 10643, 106431A. [CrossRef]
72. Soldin, R.J. SAR Target Recognition with Deep Learning. In Proceedings of the 2018 IEEE Applied Imagery Pattern Recognition
Workshop (AIPR), Washington, DC, USA, 9–11 October 2018; pp. 1–8.
73. Shahid, H.; Shah, M.A.; Almogren, A.; Khattak, H.A.; Din, I.U.; Kumar, N.; Maple, C. Machine Learning-based Mist Computing
Enabled Internet of Battlefield Things. ACM Trans. Internet Technol. 2021, 21, 1–26. [CrossRef]
74. Rosellini, A.J.; Stein, M.B.; Benedek, D.M.; Bliese, P.; Chiu, W.T.; Hwang, I.; Monahan, J.; Nock, M.K.; Sampson, N.A.;
Street, A.E.; et al. Predeployment predictors of psychiatric disorder-symptoms and interpersonal violence during combat
deployment. Depress. Anxiety 2018, 35, 1073–1080. [CrossRef]
75. Rosellini, A.J.; Monahan, J.; Street, A.E.; Heeringa, S.G.; Hill, E.D.; Petukhova, M.; Reis, B.Y.; Sampson, N.A.; Bliese, P.;
Schoenbaum, M.; et al. Predicting non-familial major physical violent crime perpetration in the US Army from administrative
data. Psychol. Med. 2016, 46, 303–316. [CrossRef]
76. Oh, D.; Han, J. Fisheye-Based Smart Control System for Autonomous UAV Operation. Sensors 2020, 20, 7321. [CrossRef]
77. Sharma, S.; Bhatt, M.; Sharma, P. Face recognition system using machine learning algorithm. In Proceedings of the 2020 5th
International Conference on Communication and Electronics Systems (ICCES), Coimbatore, India, 10–12 June 2020; pp. 1162–1168.
78. Privett, G.J.; Harvey, P.R.W.; Booth, D.M.; Kent, P.J.; Redding, N.J.; Evans, D.; Jones, K.L. Software tools for assisting the
multisource imagery analyst. In Proceedings of the Applications of Digital Image Processing XXVI, San Diego, CA, USA,
5–8 August 2003; Volume 5203, pp. 163–176. [CrossRef]
79. Martinez, J.; MacAllister, A.; Dominguez, E.; Mahlmann, C.; Quinlan, G.; McKee, M. Applying Neural Networks to the F-35 Seam
Validation Process. In Proceedings of the 2021 IEEE Aerospace Conference (50100), Big Sky, MT, USA, 6–13 March 2021; Volume
50100, pp. 1–11.
80. Jahromi, A.N.; Hashemi, S.; Dehghantanha, A.; Parizi, R.M.; Choo, K.-K.R. An Enhanced Stacked LSTM Method with No Random
Initialization for Malware Threat Hunting in Safety and Time-Critical Systems. IEEE Trans. Emerg. Top. Comput. Intell. 2020, 4,
630–640. [CrossRef]
81. Wang, X.; Cheng, P.; Liu, X.; Uzochukwu, B. Fast and Accurate, Convolutional Neural Network Based Approach for Object
Detection from UAV. In Proceedings of the IECON 2018—44th Annual Conference of the IEEE Industrial Electronics Society,
Washington, DC, USA, 21–23 October 2018; pp. 3171–3175.
82. Alfandi, O.; Khanji, S.; Ahmad, L.; Khattak, A. A survey on boosting IoT security and privacy through blockchain. Clust. Comput.
2021, 24, 37–55. [CrossRef]
83. Zhong, G.; Ling, X.; Wang, L. From shallow feature learning to deep learning: Benefits from the width and depth of deep
architectures. WIREs Data Min. Knowl. Discov. 2018, 9, 1255. [CrossRef]
84. Graham, K.; Dipnall, J.; Van Hooff, M.; Lawrence-Wood, E.; Searle, A.; Ao, A.M. Identifying clusters of health symptoms in
deployed military personnel and their relationship with probable PTSD. J. Psychosom. Res. 2019, 127, 109838. [CrossRef]
85. Dwivedi, R.K.; Pandey, S.; Kumar, R. A Study on Machine Learning Approaches for Outlier Detection in Wireless Sensor Network.
In Proceedings of the 2018 8th International Conference on Cloud Computing, Data Science & Engineering (Confluence), Noida,
India, 11–12 January 2018; pp. 189–192.
86. Arnous, F.I.; Narayanan, R.M.; Li, B. Artificial intelligence implementation for multi-domain sensor suite optimization. In
Proceedings of the Radar Sensor Technology XXIV International Society for Optics and Photonics, Bellingham, WA, USA, 27
April–8 May 2020; Volume 11408, p. 1140818. [CrossRef]
87. Zuluaga, J.G.C.; Leidig, J.P.; Trefftz, C.; Wolffe, G. Deep Reinforcement Learning for Autonomous Search and Rescue. In
Proceedings of the NAECON 2018—IEEE National Aerospace and Electronics Conference, Dayton, OH, USA, 23–26 July 2018;
pp. 521–524.
88. Shin, G.; Yooun, H.; Shin, D.; Shin, D. Incremental learning method for cyber intelligence, surveillance, and reconnaissance in
closed military network using converged IT techniques. Soft Comput. 2018, 22, 6835–6844. [CrossRef]
89. Zheng, Y.; Song, Q.; Liu, J.; Song, Q.; Yue, Q. Research on motion pattern recognition of exoskeleton robot based on multimodal
machine learning model. Neural Comput. Appl. 2020, 32, 1869–1877. [CrossRef]
90. Iftimie, I.A.; Wilson, R.L. The Use of Signals Intelligence in Offensive Cyberspace Operations: An Anticipatory Ethical Analysis.
In Proceedings of the International Conference on Cyber Warfare and Security, Norfolk, VA, USA, 12–13 March 2020; p. 262-XVIII.
[CrossRef]
Mathematics 2022, 10, 1397 27 of 27
91. Zuromski, K.L.; Bernecker, S.L.; Chu, C.; Wilks, C.R.; Gutierrez, P.M.; Joiner, T.E.; Liu, H.; Naifeh, J.A.; Nock, M.K.;
Sampson, N.A.; et al. Pre-deployment predictors of suicide attempt during and after combat deployment: Results from the Army
Study to Assess Risk and Resilience in Servicemembers. J. Psychiatr. Res. 2020, 121, 214–221. [CrossRef] [PubMed]
92. Morse, S.; Talty, K.; Kuiper, P.; Scioletti, M.; Heymsfield, S.B.; Atkinson, R.L.; Thomas, D.M. Machine learning prediction of
combat basic training injury from 3D body shape images. PLoS ONE 2020, 15, e0235017. [CrossRef]
93. Tursunbayeva, A.; Di Lauro, S.; Pagliari, C. People analytics—A scoping review of conceptual boundaries and value propositions.
Int. J. Inf. Manag. 2018, 43, 224–247. [CrossRef]

Military ML

Uploaded by

Copyright:

Available Formats

Military ML

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Military ML

Uploaded by

Copyright:

Available Formats

mathematics

1 Faculty of Statistics, Complutense University, Puerta de Hierro, 3728040 Madrid, Spain

Academic Editor: Alexander Ryzhov MSC: 68T01

Received: 28 February 2022

Mathematics 2022, 10, 1397. https://doi.org/10.3390/math10091397 https://www.mdpi.com/journal/mathematics

2.1. Previous Research on the Military Applications of ML

Table 1. Related review work.

Category of the Review Work Refs.

We have classified these works into three categories:

2.2. Background of Data-Driven Architecture for Organizations

Figure 1. Data-driven architecture for nonmilitary organizations.

3.2. Data Extraction

3.3. Data Preprocessing

3.4. Multidimensional Analysis

Table 2. Dimensions, meaning and type of analysis available.

Dimension Meaning Type of Analysis Available

sent the number of documents in which each appears.

Mathematics 2022, 10, x FOR PEER REVIEW 9 of 28

Figure 5. Example of stability between periods.

3.9. Overall Results of the Analysis

4. Application of ML in the Military Context

3.8. Visualization and Interpretation of Results

Figure 6. Preliminary network by co-occurrence.

Mathematics 2022, 10, x FOR PEER REVIEW 11 of 28

Psychological and Behavioral Disorders

ML and Opt Opt

Military Medical Studies

The areas “Electrical-Electronic Engineering”, “Computer” and “Telecommunica-

The areas “Electrical-Electronic Engineering”, “Computer” and “Telecommunications”

Table 3. Main areas.

Web of Science Categories Record Count % of 525

4.3. Country Analysis

Mathematics 2022, 10, x FOR PEER REVIEW 19 of 28

Figure 18. Network by co-authorship.

Figure 18. Network by co-authorship.

Figure 20. Network by bibliographic

4.5. Author Analysis

Figure 21. Network by citations.

4.6. Overall Results of the Analysis

Mathematics 2022, 10, x FOR PEER REVIEW 22 of 28

Figure 22. Data-driven architecture for military organizations.

6. Conclusions and Future Work

• Underlying this is an intrinsic importance in medical research that is likely to have

You might also like