Review
Advancing Maritime Safety: A Literature Review on Machine
Learning and Multi-Criteria Analysis in PSC Inspections
Zlatko Boko *, Ivica Skoko, Zaloa Sanchez Varela and Vice Milin
Faculty of Maritime Studies, University of Split, 21000 Split, Croatia; (I.S.);
(Z.S.V.); (V.M.)
* Correspondence:
Abstract: This literature review provides a structured quantitative analysis of existing re-
search on the application of machine learning models (MLMs) and multi-criteria decision-
making methods (MCDM) in the context of port state control (PSC). The aim of the study
is to capture current research trends, identify thematic priorities, and demonstrate how
these analytical tools have been used to support decision-making and risk assessment in
the maritime domain. Rather than evaluating the effectiveness of individual models, the
study focuses on the distribution and frequency of their use and provides insights into the
development of methodological approaches in this area. Although several studies suggest
that the integration of MLMs and MCDM techniques can improve the objectivity and ef-
ficiency of PSC inspections, this report does not provide a comparative assessment of their
performance. Instead, it lays the groundwork for future qualitative studies that will assess
the practical benefits and challenges of such integration. The findings suggest a frag-
mented but growing research interest in data-driven approaches to PSC and highlight the
potential of advanced analytics to support maritime safety and regulatory compliance.
Keywords: literature review; machine learning models; multi-criteria analysis; port state
control; maritime safety
Academic Editor: Xinqiang Chen
Received: 31 March 2025
Revised: 15 May 2025
Accepted: 16 May 2025 1. Introduction
Published: 17 May 2025
This paper analyzes the integration of machine learning (ML) models and multi-cri-
Citation: Boko, Z.; Skoko, I.;
teria analysis (MCA) in assessing vessel safety risks within the context of port state control
Sanchez Varela, Z.; Milin, V.
(PSC) inspections, applying a systematic literature review (SLR).
Advancing Maritime Safety:
A Literature Review on Machine The focus of this study is on a systematic literature review on the application of ad-
Learning and Multi-Criteria vanced ML and MCA methods to improve evaluation, forecasting, and planning pro-
Analysis in PSC Inspections. J. Mar. cesses in the context of PSC. The approach is primarily quantitative, with the intention of
Sci. Eng. 2025, 13, 974. https:// collecting and processing relevant literature documenting trends, scope, and specific ap-
doi.org/10.3390/jmse13050974
plications of these methods in the maritime sector. The aim is not to critically evaluate
Copyright: © 2025 by the author. each individual study or to assess the merits and limitations of particular techniques, but
Licensee MDPI, Basel, Switzerland. rather to systematically collect, categorize, and analyze existing work in order to assess
This article is an open access article
the scope, dynamics, and focus areas of current research.
distributed under the terms and
This provides an insight into the extent to which ML and MCA techniques have al-
conditions of the Creative Commons
Attribution (CC BY) license ready been integrated into inspection risk assessment, resource planning in shipping com-
(https://creativecommons.org/license panies, and the development of strategic regulatory approaches within international
s/by/4.0/). frameworks such as Memoranda of Understanding (MoU). While the practical relevance
J. Mar. Sci. Eng. 2025, 13, 974 https://doi.org/10.3390/jmse13050974
,J. Mar. Sci. Eng. 2025, 13, 974 2 of 26
of these methods for improving maritime safety is obvious, the main contribution of this
study is to take stock of existing knowledge and identify potential research gaps.
The study is part of a broader research project (PhD), with the next phase focusing
on collecting newer and more comprehensive data sources to support further investiga-
tions while strengthening the application of advanced MLMs and MCA techniques in the
context of PSC.
Maritime safety is a key component of the global transport system, and PSC [1,2] is
an integral part of the international strategy aimed at reducing the risk of marine acci-
dents. This system operates within the framework of the International Maritime Organi-
zation (IMO). It relies on regional PSC agreements, such as the Paris MoU and the Tokyo
MoU, which facilitate the exchange of information and best practices among signatory
states. PSC inspections play a crucial role in international maritime oversight, aiming to
ensure safety and the protection of life at sea, as well as the preservation of the environ-
ment from potential maritime accidents [3–6].
The main objective of PSC inspections is to identify and eliminate safety risks and
potential violations that could jeopardize the safety of the vessel, the crew, and the envi-
ronment. Based on binding international conventions such as SOLAS (Safety of Life at
Sea), MARPOL (Marine Pollution) and STCW (Standards of Training, Certification and
Watchkeeping), PSC inspectors are authorized to inspect vessels in port, regardless of the
flag under which the vessel sails, to check compliance with international standards. Dur-
ing these inspections, which may encompass various aspects of the vessel’s operation, in-
spectors assess the vessel’s condition, including its structural components, safety equip-
ment, environmental protection systems, and the qualifications of its crew. In the event of
non-compliance or irregularities, inspectors may decide to detain the vessel in port, which
can have serious consequences for both the vessel owners and its operations.
Modern risk assessment methods for PSC vessel inspections extend far beyond tra-
ditional approaches, which primarily rely on statistical analyses and expert judgment.
These are often limited by the subjective judgements of the inspectors and, from the ship-
owner’s perspective, reduce the ability to assess potential violations or the likelihood of
the vessel being detained. The use of MLMs [7,8] enables the automated processing of
large data sets [9], pattern recognition, and a more accurate determination of safety risks.
At the same time, MCA provides a systematic framework for decision-making by consid-
ering multiple relevant factors simultaneously, which improves the interpretation of the
results generated by MLMs.
The application of these methods enables a more precise prediction of inspection re-
sults. It is therefore extremely useful for both inspectors, who need to formulate require-
ments by legal standards, and shipowners, who want to minimize costs and optimize the
time vessels spend in port. By utilizing advanced technologies, including artificial intelli-
gence (AI), machine learning (ML), and algorithms, it is possible to optimize procedures
for identifying potential risks based on historical data on violations and deficiencies on
vessels. This approach not only improves the efficiency of inspections but also helps re-
duce the number of accidents caused by technical defects [10–13].
In this context, current research [14–19] focuses on the development of sophisticated
MLMs that enable automatic pattern recognition in data from previous inspections and a
more reliable prediction of future violations and safety risks. The main challenges in the
field of PSC inspections and vessel safety are to increase the accuracy of predictions and
improve the robustness of these models to ensure their applicability in real operational
conditions.
The effectiveness of different ML and multi-criteria decision-making (MCDM) meth-
ods in PSC depends on the specific requirements of the inspection process. Methods such
as decision trees are highly interpretable and easy for inspectors to understand, but often
, J. Mar. Sci. Eng. 2025, 13, 974 3 of 26
have lower accuracy in more complex scenarios. In contrast, methods such as Random
Forests (RFs) and Support Vector Machines (SVMs) offer higher predictive performance
and accuracy, but their complexity reduces transparency and makes the results more dif-
ficult to interpret. Bayesian network (BN) algorithms are fast and simple and are therefore
suitable for initial analyses, but are limited by the assumption of feature independence.
On the other hand, MCDM methods such as the Analytic Hierarchy Process (AHP) and
the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) allow a
structured ranking of vessels based on risk and other factors, which support informed
decision-making, although they rely on subjective judgments that may influence the final
outcome. In practice, an optimal approach often requires the combination of interpretable
and powerful methods to find a balance between the reliability of the model and the con-
fidence of the inspectors.
The combination of MLMs and MCA opens up the possibility of improving the effi-
ciency of PSC inspections and increasing maritime safety [20–23].
The aim of this paper is therefore to conduct a SLR for the period of the last ten years
(from 2015 to 2025) to analyze existing approaches to the integration of MLMs and MCA
in the assessment of vessel safety risks in the context of PSC reviews, to identify essential
methods and research gaps, and to define possible directions for future research.
The paper is organized in such a way that the introductory section presents the basic
concepts of the research, focusing on the importance of applying MLMs and MCA in as-
sessing safety risks within the context of PSC inspections. The Materials and Methods chap-
ter outlines the framework of the research methodology, including the search and analysis
strategy for relevant literature, the criteria for selecting sources, and the data used for
analysis. The Results and Discussion chapters present the analysis results, along with a
discussion of the key findings and challenges encountered in implementing MLMs and
MCA methods within the context of PSC inspections. Finally, the Conclusions chapter sum-
marizes the main findings of the study. It formulates recommendations for future research
and improvement of the methodological approach to safety risk assessment through the
application of modern technological solutions and their implementation in the field of
PSC inspections.
1.1. Maritime Safety and PSC Inspection
In the section “Maritime Safety and PSC Inspection”, the studies analyzed highlight the
most important aspects of maritime safety and the role of PSC inspections in maintaining
this safety. The main findings of these studies relate to the importance of PSC inspections,
the impact of the COVID-19 pandemic in this area, the most commonly identified defi-
ciencies in PSC inspections, human factors as key elements of maritime safety, and the
researchers’ emphasis on the need for technological improvements and better regulations.
The data collected during PSC inspections form the basis for improving safety in the
maritime industry and enable the development of strategies to improve inspection pro-
cesses and predict potential risks on vessels [24,25].
Considering that PSC inspections are one of the most important mechanisms for en-
suring maritime safety and reducing operational risks, their role in identifying violations
on vessels is crucial, particularly in contributing to the reduction of marine casualties and
environmental incidents.
Studies [26–29] confirm that PSC inspections play a key role in the implementation
of international regulations such as SOLAS and MARPOL and ensure a high level of safety
worldwide. Their effectiveness in reducing the risk of accidents through preventive
measures and penalizing vessels that do not meet safety standards is particularly empha-
sized.
Advancing Maritime Safety: A Literature Review on Machine
Learning and Multi-Criteria Analysis in PSC Inspections
Zlatko Boko *, Ivica Skoko, Zaloa Sanchez Varela and Vice Milin
Faculty of Maritime Studies, University of Split, 21000 Split, Croatia; (I.S.);
(Z.S.V.); (V.M.)
* Correspondence:
Abstract: This literature review provides a structured quantitative analysis of existing re-
search on the application of machine learning models (MLMs) and multi-criteria decision-
making methods (MCDM) in the context of port state control (PSC). The aim of the study
is to capture current research trends, identify thematic priorities, and demonstrate how
these analytical tools have been used to support decision-making and risk assessment in
the maritime domain. Rather than evaluating the effectiveness of individual models, the
study focuses on the distribution and frequency of their use and provides insights into the
development of methodological approaches in this area. Although several studies suggest
that the integration of MLMs and MCDM techniques can improve the objectivity and ef-
ficiency of PSC inspections, this report does not provide a comparative assessment of their
performance. Instead, it lays the groundwork for future qualitative studies that will assess
the practical benefits and challenges of such integration. The findings suggest a frag-
mented but growing research interest in data-driven approaches to PSC and highlight the
potential of advanced analytics to support maritime safety and regulatory compliance.
Keywords: literature review; machine learning models; multi-criteria analysis; port state
control; maritime safety
Academic Editor: Xinqiang Chen
Received: 31 March 2025
Revised: 15 May 2025
Accepted: 16 May 2025 1. Introduction
Published: 17 May 2025
This paper analyzes the integration of machine learning (ML) models and multi-cri-
Citation: Boko, Z.; Skoko, I.;
teria analysis (MCA) in assessing vessel safety risks within the context of port state control
Sanchez Varela, Z.; Milin, V.
(PSC) inspections, applying a systematic literature review (SLR).
Advancing Maritime Safety:
A Literature Review on Machine The focus of this study is on a systematic literature review on the application of ad-
Learning and Multi-Criteria vanced ML and MCA methods to improve evaluation, forecasting, and planning pro-
Analysis in PSC Inspections. J. Mar. cesses in the context of PSC. The approach is primarily quantitative, with the intention of
Sci. Eng. 2025, 13, 974. https:// collecting and processing relevant literature documenting trends, scope, and specific ap-
doi.org/10.3390/jmse13050974
plications of these methods in the maritime sector. The aim is not to critically evaluate
Copyright: © 2025 by the author. each individual study or to assess the merits and limitations of particular techniques, but
Licensee MDPI, Basel, Switzerland. rather to systematically collect, categorize, and analyze existing work in order to assess
This article is an open access article
the scope, dynamics, and focus areas of current research.
distributed under the terms and
This provides an insight into the extent to which ML and MCA techniques have al-
conditions of the Creative Commons
Attribution (CC BY) license ready been integrated into inspection risk assessment, resource planning in shipping com-
(https://creativecommons.org/license panies, and the development of strategic regulatory approaches within international
s/by/4.0/). frameworks such as Memoranda of Understanding (MoU). While the practical relevance
J. Mar. Sci. Eng. 2025, 13, 974 https://doi.org/10.3390/jmse13050974
,J. Mar. Sci. Eng. 2025, 13, 974 2 of 26
of these methods for improving maritime safety is obvious, the main contribution of this
study is to take stock of existing knowledge and identify potential research gaps.
The study is part of a broader research project (PhD), with the next phase focusing
on collecting newer and more comprehensive data sources to support further investiga-
tions while strengthening the application of advanced MLMs and MCA techniques in the
context of PSC.
Maritime safety is a key component of the global transport system, and PSC [1,2] is
an integral part of the international strategy aimed at reducing the risk of marine acci-
dents. This system operates within the framework of the International Maritime Organi-
zation (IMO). It relies on regional PSC agreements, such as the Paris MoU and the Tokyo
MoU, which facilitate the exchange of information and best practices among signatory
states. PSC inspections play a crucial role in international maritime oversight, aiming to
ensure safety and the protection of life at sea, as well as the preservation of the environ-
ment from potential maritime accidents [3–6].
The main objective of PSC inspections is to identify and eliminate safety risks and
potential violations that could jeopardize the safety of the vessel, the crew, and the envi-
ronment. Based on binding international conventions such as SOLAS (Safety of Life at
Sea), MARPOL (Marine Pollution) and STCW (Standards of Training, Certification and
Watchkeeping), PSC inspectors are authorized to inspect vessels in port, regardless of the
flag under which the vessel sails, to check compliance with international standards. Dur-
ing these inspections, which may encompass various aspects of the vessel’s operation, in-
spectors assess the vessel’s condition, including its structural components, safety equip-
ment, environmental protection systems, and the qualifications of its crew. In the event of
non-compliance or irregularities, inspectors may decide to detain the vessel in port, which
can have serious consequences for both the vessel owners and its operations.
Modern risk assessment methods for PSC vessel inspections extend far beyond tra-
ditional approaches, which primarily rely on statistical analyses and expert judgment.
These are often limited by the subjective judgements of the inspectors and, from the ship-
owner’s perspective, reduce the ability to assess potential violations or the likelihood of
the vessel being detained. The use of MLMs [7,8] enables the automated processing of
large data sets [9], pattern recognition, and a more accurate determination of safety risks.
At the same time, MCA provides a systematic framework for decision-making by consid-
ering multiple relevant factors simultaneously, which improves the interpretation of the
results generated by MLMs.
The application of these methods enables a more precise prediction of inspection re-
sults. It is therefore extremely useful for both inspectors, who need to formulate require-
ments by legal standards, and shipowners, who want to minimize costs and optimize the
time vessels spend in port. By utilizing advanced technologies, including artificial intelli-
gence (AI), machine learning (ML), and algorithms, it is possible to optimize procedures
for identifying potential risks based on historical data on violations and deficiencies on
vessels. This approach not only improves the efficiency of inspections but also helps re-
duce the number of accidents caused by technical defects [10–13].
In this context, current research [14–19] focuses on the development of sophisticated
MLMs that enable automatic pattern recognition in data from previous inspections and a
more reliable prediction of future violations and safety risks. The main challenges in the
field of PSC inspections and vessel safety are to increase the accuracy of predictions and
improve the robustness of these models to ensure their applicability in real operational
conditions.
The effectiveness of different ML and multi-criteria decision-making (MCDM) meth-
ods in PSC depends on the specific requirements of the inspection process. Methods such
as decision trees are highly interpretable and easy for inspectors to understand, but often
, J. Mar. Sci. Eng. 2025, 13, 974 3 of 26
have lower accuracy in more complex scenarios. In contrast, methods such as Random
Forests (RFs) and Support Vector Machines (SVMs) offer higher predictive performance
and accuracy, but their complexity reduces transparency and makes the results more dif-
ficult to interpret. Bayesian network (BN) algorithms are fast and simple and are therefore
suitable for initial analyses, but are limited by the assumption of feature independence.
On the other hand, MCDM methods such as the Analytic Hierarchy Process (AHP) and
the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) allow a
structured ranking of vessels based on risk and other factors, which support informed
decision-making, although they rely on subjective judgments that may influence the final
outcome. In practice, an optimal approach often requires the combination of interpretable
and powerful methods to find a balance between the reliability of the model and the con-
fidence of the inspectors.
The combination of MLMs and MCA opens up the possibility of improving the effi-
ciency of PSC inspections and increasing maritime safety [20–23].
The aim of this paper is therefore to conduct a SLR for the period of the last ten years
(from 2015 to 2025) to analyze existing approaches to the integration of MLMs and MCA
in the assessment of vessel safety risks in the context of PSC reviews, to identify essential
methods and research gaps, and to define possible directions for future research.
The paper is organized in such a way that the introductory section presents the basic
concepts of the research, focusing on the importance of applying MLMs and MCA in as-
sessing safety risks within the context of PSC inspections. The Materials and Methods chap-
ter outlines the framework of the research methodology, including the search and analysis
strategy for relevant literature, the criteria for selecting sources, and the data used for
analysis. The Results and Discussion chapters present the analysis results, along with a
discussion of the key findings and challenges encountered in implementing MLMs and
MCA methods within the context of PSC inspections. Finally, the Conclusions chapter sum-
marizes the main findings of the study. It formulates recommendations for future research
and improvement of the methodological approach to safety risk assessment through the
application of modern technological solutions and their implementation in the field of
PSC inspections.
1.1. Maritime Safety and PSC Inspection
In the section “Maritime Safety and PSC Inspection”, the studies analyzed highlight the
most important aspects of maritime safety and the role of PSC inspections in maintaining
this safety. The main findings of these studies relate to the importance of PSC inspections,
the impact of the COVID-19 pandemic in this area, the most commonly identified defi-
ciencies in PSC inspections, human factors as key elements of maritime safety, and the
researchers’ emphasis on the need for technological improvements and better regulations.
The data collected during PSC inspections form the basis for improving safety in the
maritime industry and enable the development of strategies to improve inspection pro-
cesses and predict potential risks on vessels [24,25].
Considering that PSC inspections are one of the most important mechanisms for en-
suring maritime safety and reducing operational risks, their role in identifying violations
on vessels is crucial, particularly in contributing to the reduction of marine casualties and
environmental incidents.
Studies [26–29] confirm that PSC inspections play a key role in the implementation
of international regulations such as SOLAS and MARPOL and ensure a high level of safety
worldwide. Their effectiveness in reducing the risk of accidents through preventive
measures and penalizing vessels that do not meet safety standards is particularly empha-
sized.
