Reshaping 3PL Operations: Machine Learning Approaches to Mitigate
and Manage Damage Parameters
YUNUS EMRE YETİŞ, SAFIYE TURGAY, BİLAL ERDEMİR
Department of Industrial Engineering,
Sakarya University,
54187, Esentepe Campus Serdivan-Sakarya,
TURKEY
Abstract: - In the third-party logistics (3PL) environment, it is very important to reduce damage parameters,
increase operational efficiency and reduce costs. This study aims to develop strategies for reshaping 3P
operations by analyzing the parameters involved in damage control with machine learning. The logistics sector
is gradually growing in the world and the potential of the sector is better understood over time. Damage to
products in the logistics sector, especially during transportation and storage, not only causes financial losses but
also affects customer productivity and operational efficiency. With the use of artificial intelligence techniques,
it is possible to determine consumer expectations, predict damage losses, and develop innovative strategies by
applying machine learning algorithms. At the same time, options such as driverless vehicles, robots used in
storage and shelves, and the easy use of big data within the system, which have emerged with artificial
intelligence, minimize errors in the logistics sector. Thanks to the use of artificial intelligence in the logistics
sector, businesses are more efficient. This study includes an estimation study in the field of error parameters for
the logistics service sector with machine learning methods. In the application, real data of a 3PL company for
the last 5 years is used. For the success of 3PL companies, warehousing and undamaged delivery of products
are of great importance. The fewer damaged products they send, the more they increase their value. The
company examined in the study kept its damage data and wanted it to be analyzed so that it could take
precautions accordingly and follow a more profitable path. For this reason, the study focuses on data on errors
and damages. This study shows what kind of problems can occur in such a company and how the 3PL company
can evaluate the problems to increase customer service quality and cost efficiency.
Key-Words: - Third Party Logistics (3PL), Machine learning, Damage Parameters, Supply Chain Optimization,
Predictive Analytics, IoT, Big Data, AzureML.
Received: August 11, 2023. Revised: December 7, 2023. Accepted: February 13, 2024. Published: April 4, 2024.
1 Introduction
Third-party logistics (3PL) faces the key challenges
of achieving operational efficiency, reducing risk,
and increasing customer satisfaction in dynamic and
complex environments. Effective management and
mitigation of damage parameters, focusing on
management, and maintaining uninterrupted supply
chain processes at the same time are of vital
importance. 3PL aims to develop proactive and
data-driven solutions to ensure the delivery of
products in a global environment, the difficulties
encountered in logistics operations and to overcome
these difficulties.
The machine learning program used is
Microsoft Machine Learning Studio. The data
obtained from the company is brought to csv format
and processed in the program. Regression,
clustering, and classification studies are carried out
here and it is examined which one gives more
effective results.
This study aimed to analyze historical data to
reveal the relationship and identify risk factors,
predict potential damage scenarios and develop
predictive models, and use machine learning
algorithms to minimize damages and ensure overall
operational resilience in 3PL operations. As can be
seen in the results, the data set used was not suitable
for regression and did not give good enough results.
However, because of the algorithms used and the
applications made, it was observed that clustering
and classification results gave good results. Answers
to questions such as which error leads to which
consequences, for which reasons mistakes are made,
for which reasons and mistakes the most damage
bills are issued, etc. have been revealed. The
company that examines these can easily understand
and decide where to take precautions and work. In
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
12
Volume 23, 2024
this study, regression, classification, and clustering
algorithms are used to predict and categorize
potential risk areas.
The main objective of this study is to minimize
the impact on supply chain management with the
help of machine learning techniques by analyzing
the damage parameters in the 3PL framework.
Using historical data of the 3PL company, strategies
were developed to identify risk factors, and predict
and prevent damages. With data-driven insights,
predictive analytics, and damage mitigation
approaches, 3PL operations are able to reshape and
more resilient and efficient strategies developed.
The remainder of this paper is structured as
follows: Section 2 includes an overview of the
pertinent literature. Section 3 defines the model
definition and formulation, encompassing the
mathematical model of the reshaping 3PL
operations. Section 4 structures a case study,
summarizing the key results of our proposed
approach in comparison to the current state of
affairs. Finally, in Section 5, we present our
concluding remarks.
2 Literature Survey
3PL industry, facilitating the global movement of
products and minimizing damage are among the
main objectives. Predicting the main complexities
and challenges encountered in logistics operations
and taking the necessary preventive measures are
among the main objectives of this study. In this
context, the studies in the literature examined to
maintain uninterrupted logistics management. In
these studies, the historical development of 3PL
strategies was followed and the transition process
from traditional approaches to modern and
technology-oriented solutions was examined, [1],
[2], [3], [4], [5], [6].
Optimizing routes, predicting customer trends,
[7], [8], [9] and establishing the balance between the
capacities of warehouses and customer demands,
[10], [11], have been evaluated with machine
learning concepts in logistics. Along with AI
techniques, the use of IoT and blockchain has
become widespread, [12], [13], [14], [15], [16].
Some applied studies have been done to evaluate
dynamic service states, [17], [18], [19], [20], [21],
[22], [23]. Theoretical and applied studies
evaluating the performance of the system structure
are available in the literature, [24], [25], [26], [27],
[28], [29], [30], [31].
Reshaping Third-Party Logistics (3PL)
operations through the application of machine
learning approaches to mitigate and manage damage
parameters in the supply chain is a dynamic and
innovative area. Some of the examples are:
- Damage Prediction in Transit
Consider a 3PL provider responsible for
transporting fragile goods. Machine learning models
are trained on historical data to predict the
likelihood of damage during transit based on factors
such as transportation mode, route conditions, and
weather. By using machine-learning algorithms for
damage prediction, 3PLs can proactively identify
high-risk routes or transportation methods. This
allows for better planning, route optimization, and
the implementation of preventive measures,
ultimately minimizing the occurrence of damages,
[32], [33], [34].
- Package and Handling Optimization
Machine learning algorithms can analyze data
related to packaging materials, handling procedures,
and transportation methods to optimize packaging
configurations and reduce the risk of damage.
Research in this context can assess the effectiveness
of machine learning in recommending optimal
packaging solutions. By understanding the
relationships between packaging attributes and
damage incidents, 3PLs can implement data-driven
improvements in the way packages are prepared and
handled.
- Real-time Monitoring with IoT
Enabling the joint operation of machine learning
with IoT devices in monitoring the physical
environment (such as the temperature, and
humidity) as well as vibrations during shipment is
the key feature of this system. Research may include
creating machine learning models for analysis of
streaming data coming from IS devices which can
be used for anomaly or deviation-seeking purposes
from functional sites. Immediate actions are
initiated as soon as the control system detects an
environmental parameter that violates safe limits,
consequently harming sensitive goods.
- Predictive Maintenance for Handling
Equipment
Machine learning algorithms can forecast
maintenance requirements for the handling
equipment (e.g., forklifts, conveyor belts;) by
examining historical data on resources and engine
sensor data. Research works can provide an
evaluation condition of machine learning models in
predicting equipment fails or malfunctions. Through
preventive maintenance, 3PLs can halt unnecessary
dismantling of equipment before breakdowns and
consequently be timely in their operations without
risking damages.
- Claims Assessment and Fraud Detection
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
13
Volume 23, 2024
Machine learning algorithms are used to read claim
data and find patterns that relate to cases of
fraudulent claims or damaged wishes deliberately. -
Research may unveil the efficiency of artificial
intelligence in the area of claims assessment and
proceeding process automation. Providing for the
distinction between authentic claims and malicious
activities, 3PLs will enable the fast resolution of
claims, decrease penalties, and curb fraudulent
actions.
- Supplier and Carrier Performance Analysis
Through this approach, machine learning can find
the relations between suppliers' or carriers practices
and eventual problems. The study agenda was
narrowed down and tailored toward uncovering the
role of machine learning in supply chain
performance measuring. Through the selection of
partners who have low deteriorating rates, 3PLs can
make the correct collaboration decisions and the
overall supply chain resilience becomes more
improved.
- Customized Routing for Fragile Goods
The application of ML algorithms can generate
personalized routes for freighting perishable goods
using historical information, present road
conditions, and known risks. The evaluation of the
impact of route tailoring is used to study the
possibility of reducing the environmental
damage. Using optimized transport systems of 3PLs
with cargo particularities in mind. It is possible to
make sensitive shipments even safer and reduce
risks arising from them as well.
Machine learning methods used in the
processing of damage parameters in Third-Party
Logistics (3PL) operations have proved useful over
time, providing a better alternative that is more
advanced and data-driven for the management of
supply chain logistics. We can use machine-learning
algorithms that not only memorize historical data
and current relevant information but also can extract
sophisticated patterns, which people can hardly
recognize by means of traditional ways of obtaining
knowledge. The traditional route generally is a rule-
based system or static decision-making and thus
always lacks the talent to predict and adapt to
machine learning. Machine learning algorithms can
adjust to current environmental conditions as well as
all the technological advances such as in handling
procedures, and transportation networks by
regularly learning from newly fed data. The old-
fashioned way might be quite resistant to
unexpected events and very tiring in the case of a
scenario change. Machine learning algorithms are
best at pattern and relationship detection across
large datasets. Such a more subtle approach enables
to be able to identify the factors that lead to the
damages. There are some older models, which
mostly depend on the simple and generalized
principle and may fail to reflect the relationship
patterns in big data. Customizing a machine-
learning model to a specific logistics environment
involves considering the uniqueness of the
environmental features of the products, handling
equipment, and transportation modes. By definition,
conventional approaches should have standardized
principles, which may be out of order for logistics
operations' invariable nature.
Machine learning models integrated with IoT
devices render such operations as real-time
monitoring and decision-making possible, directly
leading to prompt responses in the case of deviants
from the optimal temperature. Traditional feature
seasons or static law models may not be able to
react promptly to nuanced situations as dynamic
machine learning algorithms. Machine learning
models are continually capable of learning and
advancing their forecasts thanks to the increasing
availability of data, hence their attunement to the
adaptive nature of logistical problems. Traditional
approaches are usually tied to manual updates and
revisions. It does not engage in self-learning and
continuing development without any human
help. Machine learning-based machine automates
disaster prediction, assessment, and decision-
making procedures, which gives an edge in the
context of logistic efficiency. Old techniques might
mean physical evaluations, judgments, and even
separate decision-making that might be a fuzzy
process. Machine learning models may process and
analyze multiple data sources, which consist of
historical records, real-time sensor data, and
environmental conditions. This enables us to have a
whole picture for logistics operations. The
conventional techniques may be constrained by the
nature of the data utilized and may not unlock the
value of collected data. Machine learning models
can scale efficiently to handle large and complex
datasets, making them suitable for 3PL operations
with varying scales and complexities. Traditional
methods may face scalability challenges, especially
when dealing with extensive and dynamic logistics
networks.
This study aims to increase the efficiency of the
overall supply chain structure by analyzing the
routes at the same time with machine learning
algorithms. In predictive modeling for analyzing the
efficiency process, past data is taken and analyzed
to make strategic decisions about the future. This
process involves the prediction of possible
disruptions. At the same time, some studies have
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
14
Volume 23, 2024
addressed the difficulties in applying machine
learning algorithms in logistics.
In summary, machine-learning approaches offer
advantages in terms of predictive capabilities,
adaptability, complex pattern recognition,
customization, real-time decision-making,
continuous learning, efficiency, and scalability.
While traditional methods have their merits, the
data-driven and dynamic nature of machine learning
makes it a valuable and advantageous tool for
reshaping 3PL operations and mitigating damage
parameters in supply chain logistics.
In this study, the machine-learning algorithm is
applied to the logistics dataset to determine the
difficulties encountered and to evaluate the
situations that may be encountered in the future.
3 Model Definition and Formulation
Reshaping third-party logistics (3PL) operations
through the integration of machine learning
approaches to reduce and manage damage
parameters represents a significant leap forward in
improving the efficiency, accuracy, and overall
performance of logistics processes. Throughout this
research, we have explored various aspects of how
machine learning can be applied to minimize and
cope with damages in the logistics industry.
In this section, the applied method and process
steps are given. The methodology and process steps
are given in Figure 1. We collected historical data
on the logistics operations of the 3PL company,
including shipments, transportation routes, handling
procedures, and damage sampling information. The
data and prediction values in the data set were
analyzed along with the efficiency cases for strategy
formulation. In the data preprocessing process,
preprocessing operations are performed by cleaning
the data, normalization, and editing missing values.
It is very important to extract relevant information
and select relevant attributes. With the relevant
information, the variables to analyze in machine
learning are determined. In this study, the
distribution of damage parameters, information
acquisition, identification of patterns, and potential
correlations analyzed. In feature selection, machine
learning algorithms use statistical methods to
identify variables that are sensitive to the data
attribute dimension. Therefore, regression models,
classification, and clustering algorithms are
investigated to predict damages in 3PL operations.
In the training and validation process, the results are
compared between the training data set and the
validation data set in the data set used, providing
realistic generalization about the invisible parts of
the models. The selection of hyperparameters is
performed which will reveal the prediction
accuracy. Accuracy, precision, and mean squared
error are used to evaluate the performance of the
proposed model. Depending on the results obtained,
it aimed to monitor high-risk shipments in real-time
to formulate strategies for damage mitigation (in
Figure 1).
In this context, this study not only estimates the
damage parameters in 3PL operations but also uses
machine learning algorithms for proactive damage
mitigation.
Fig. 1: Suggested Model
3.1 Mathematical Programming
In this model, an optimization model was designed
by taking into account the variables and constraints
determined by the mathematical model in the
forecasting of 3PL operations. The designed model,
it is aimed to minimize 3PL damages and maximize
supply chain flexibility. The decision variables in
the proposed model are packaging choices and
processing protocols. With this model, it is also
aimed to find the appropriate values for the
variables. The constraints in the model include
capacity constraints, time constraints, budget
constraints, and logistical and operational
constraints related to the 3PL environment. With the
machine learning algorithms of this model, routing
decisions can be made dynamically in high-risk
situations with machine learning predictions.
Objective Function: The objective is to minimize
the total damage in routing, packaging, and
transportation activity decisions. The function is;
(1)
Z is the total expected damages to be minimized.
is the predicted damage probability for shipment
i on route j.
is a binary decision variable indicating whether
shipment i is assigned to route j.
Decision Variables:
{0,1}: Binary decision variable representing
whether shipment i is assigned to route j
Constraints
1. Routing Constraints:
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
15
Volume 23, 2024
a. Ensure each shipment is assigned to exactly
one route:
(2)
b. Ensure routes are within capacity limits:
(3)
2. Handling and Packaging Constraints:
Integrate machine learning insights to
dynamically adjust handling and packaging
decisions based on predicted damage
probabilities.
3. Budget Constraints:
a. Limit the total cost associated with routing,
handling and packaging decisions:
(4)
Integration of machine learning algorithms into
the proposed model is very important in 3PL Loss
mitigation; after the uncertainty modeling and
determination of the parameters affecting the
system; a real-time decision support model designed
by using machine-learning algorithms to adapt the
system to changing conditions within the scope of
real-time adaptability.
Random Forest method for damage estimation
used to estimate the probability of damage for each
shipment on different routes and to determine the
relationship between the data. A logistic regression
method for route assignment is used to determine
the optimal assignment of shipments to different
routes according to the estimated damage
probabilities. The binary classification approach
provides an auxiliary framework for the decision
problem. K-Means clustering is used to evaluate
groups of strategies to help customize the handling
processes of different types of shipments. It also
uses feature importance analysis XGBoost and a
gradient boosting algorithm for damage prediction.
At the same time, depending on the dynamic
structure of the system, in the continuous
improvement process, the system constantly renews
itself according to the changing conditions with the
feedback obtained from the model with the
reinforcement learning method and completes the
continuous improvement process with reinforcement
learning (Figure 2).
System performance will be improved through
predictive modeling, optimization in 3PL logistics
management, damage parameter reduction and
management, and continuous learning.
Fig. 2: Machine Learning-Based Model Steps
4 Case Study
This study aims to proactively reduce and manage
the damage parameters that may occur during
transportation and increase operational efficiency. A
third-party logistics (3PL) provider embarked on a
transformative journey to reshape its operations
using machine learning approaches. The goal was to
proactively reduce and manage damage parameters,
increase operational efficiency, and improve overall
service quality. By taking into account the damage
estimation and route assignment situations, the
allocation of shipments to routes is done by taking
into account the probability situations.
The problem addressed in the study is to
examine and analyze the causes and frequency of
damage to minimize the damage caused by the
damages of the 3PL company due to different
reasons in recent years. In this process, the available
data was first analyzed in Excel environment and
then converted to CSV format. The file was then
uploaded to AzureML. Fault causes were analyzed
by classification method. Damage amounts were
estimated by regression method and strategic
decisions were obtained.
4.1 Analysis of Error Causes by
Classification Method
Firstly, the Excel file is converted to CSV format in
order to work in the Azure application.
Fig. 3: Azure Module Screen out
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
16
Volume 23, 2024
The data set converted from this section is loaded
into the system (Figure 3, Figure 4 and Figure 5).
Fig. 4: Azure dataset screen
From the Saved dataset section, drag the design
dataset from my dataset section to the screen.
Fig. 5: Azure dataset module
Drag the Select columns in the dataset module
and select the columns you want to work on ( Figure
6 and Figure 7).
Fig. 6: Drug the select column screen for data
The launch column selector section is presented
in Figure 7.
Fig. 7: Column section
Fault causes, faults, and products columns are
selected (Figure 8).
Fig. 8: Features analysis screen
The reason why these columns are preferred is
that it is desired to see which product is associated
with which error and for what reason (Figure 9).
Fig. 9: Edit Screen
Edit metadata module is added and displayed in
Figure 10.
Fig. 10: Edit data module
Data is organized categorically (Figure 11).
Fig. 11: Data organization module
With the Split data module, the model is set to 80%
training and 20% testing (Figure 12).
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
17
Volume 23, 2024
Fig. 12: Split data module
Then the main part can be started. Select the
Multiclass Logistic Regression module. No change
is made in the Properties section (Figure 13).
Fig. 13: Train Data Module
In the Train model option, the error causes
column is selected. The reason is that multiple faults
are linked to a single fault cause. In this way, a
clearer and cleaner classification and analysis can be
made (Figure 14).
Fig. 14: Cleaner classification and analysis module
The same process is done for the Multiclass
Neural Network method (Figure 15 and Figure 16).
Fig. 15: Multiclass Neural Network module
As a result, such a model is built. We look at the
outputs and examine which module has performed
better classification for the data set.
This is the output when the Evaluate model
section is visualized (Figure 16).
Fig. 16: Data visualization module
When this section is examined, it is seen that the
classification model made with the Multiclass
Neural Network module gives a better result. It is
seen how many percent of the error causes cover the
errors. The model where the error cause and error
matching are more accurate is the Neural model.
That process is preferred.
4.2 Analysis of Errors with Clustering
Method
The data set loaded into the system in Figure 17 is
brought to the screen by following the steps in
Figure 18 and Figure 19.
Fig. 17: Edit screen
Select the Edit metadata module and change the
desired columns to categorical.
Fig. 18: Select column screen
Select columns in the dataset module selects the
desired columns.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
18
Volume 23, 2024
Fig. 19: Split data screen
A split data module is used to set the training
and test datasets (Figure 20).
Fig. 20: Split data analysis screen
The K-Means clustering module is selected and
a dataset of 31 data objects is partitioned into 9
clusters given as input parameters (Figure 21).
Fig. 21: K-Means clustering screen
The reason for this is to be able to evolve from a
column of faults with 31 variables to a column of
fault causes with 9 variables.
The faults parameter is selected with the Train
clustering model module in Figure 22.
Fig. 22: Select column screen
Assigning data to clusters module is selected to
assign data to clusters. The columns and clusters to
be assigned data are selected with the selected
columns in the dataset module (Figure 23).
This allows us to see how much data the error
causes, errors, and assignments options have.
Fig. 23: Analysis process screen
When the final model is visualized as shown in
the Figure 24 and Figure 25.
Fig. 24: Faults and Fault reasons
Such a result is obtained. As can be seen from
the figure, the faults are assigned to certain clusters
and it is seen how often and for what reasons these
faults occur.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
19
Volume 23, 2024
Fig. 25: Fault reasons and assignments
When the error causes and clusters are analyzed
in the figure, it is seen how many numbered errors
lead to which cause and how much (Figure 26).
Fig. 26: Faults frequencies
Figure 27 shows how many values the clusters
contain (since each cluster contains one error, the
number of errors is also shown).
Fig. 27: Risk factors and risk reasons distribution
Another analysis shows which error is attributed
to which cause and how much, with the y-axis being
the error causes and the x-axis being the errors.
4.3 Estimation of Damage Amounts by
Regression Method
The stages are repeated (Figure 28).
Fig. 28: Damage estimation screen
Select the Edit metadata module and change the
damage type column from numeric to integer. And
categorical (Figure 29 and Figure 30).
Fig. 29: Illustration of damage type
Fig. 30: Clean missing data screen
Clean missing data clears the rows with missing
data in the columns we will use in Figure 31.
Fig. 31: Data selection screen
Select columns in the dataset to select the
columns to work on (Figure 32).
Fig. 32: Normalization data screen
The data set, which is normalized to MinMax
format with normalized data, is separated into
training and testing with split data. Then the linear
regression module is connected to the set (Figure
33).
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
20
Volume 23, 2024
Fig. 33: Tune model screen
The tune model module is used to select the
invoice amount to be estimated and the score model
and then the evaluate model modules are connected
and the training of the model is finished (Figure 34).
Fig. 34: Finished off the training screen
The outputs of the model, the final version of
which are analyzed in Table 1.
Table 1. Analysis results
As can be seen, regression does not give very
good results. In the model tested with various
regression methods, linear regression gave the best
result, but it was not good enough.
The results of the study are given in Figure 25,
Figure 26, Figure 27, Figure 29 and Table 1. As can
be seen from these results, good and useful results
were obtained with the classification and clustering
methods. Regression was not very reliable and
useful for this data set. Large differences were
observed between the predictions and the actual
results.
5 Conclusion
As a result, the study has provided the company
with information about where it is experiencing
problems and what measures it needs to take, what it
needs to change, or what it needs to disable.
Machine learning models, including predictive
analytics, computer vision, and anomaly detection,
offer valuable tools for predicting, identifying, and
preventing damage during the handling and
transportation of goods. Through data patterns and
anomaly recognitions from historical data, these
models provide logistics companies a prior base to
deal with critical problems whereas they may
worsen.
In addition, digital technologies that bring in
real-time monitoring systems with sensors and IoT
devices remarkably increase the traceability and
visibility of goods during the supply chain. To this
end, AI can help identify potential damage on time
and supply the data that is used to improve
processes and logistics optimization.
What is more, artificial intelligence-powered
algorithms are able to do the optimal route planning
and warehouse management, which, in turn, will
ensure that goods are transported safely and
efficiently with the lesser risk of damage. These
algorithms consider such diverse factors as traffic,
weather, and, in addition, the fragility of goods to
provide the necessary changes for logistics
processes in real-time.
Along with the introduction of machine learning
in the processes of 3PL and the administration
services, it is of major significance to pay attention
to such problems as the quality of data, model
interpretability, and scalability. Companies have to
spend on resilient data management practices,
operational model management, upgrade and routine
check to keeping pace with the emerging dynamics
of logistics.
In the end, we can say that the creation of ML in
three-player operator machine is not just a
technological upgrade. It is a competitive strategy of
companies, which are willing to get an advantage in
the quickly changing and growing logistics
landscape. The new chapter of AI and the logistics
industry might appear in its innovative solutions to
the problem of freight loss, operational efficiency,
and customer satisfaction.
Implementing these innovations will not only
reduce costs per unit and drive profitability, but it is
also a way of expressing that logistics companies
can meet supply chain needs for reliable and
resilient supply chain solutions.
References:
[1] Duong, Q.H., Zhou, L., Meng, M., Nguyen,
T.V., Ieromonachou, P., Nguyen, D.T.,
Understanding product returns: A systematic
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
21
Volume 23, 2024
literature review using machine learning and
bibliometric analysis, International Journal of
Production Economics, Vol. 243, January
2022, 108340.
[2] Sun, Y., Kirtonia, S., Chen, Z.L., A survey of
finished vehicle distribution and related
problems from an optimization perspective,
Transportation Research Part E: Logistics
and Transportation Review, Vol. 149, May
2021, 102302
[3] Golob, T.F., Regan, A.C., Impacts of
information technology on personal travel and
commercial vehicle operations: research
challenges and opportunities, Transportation
Research Part C: Emerging Technologies,
Vol. 9, Issue 2, April 2001, Pages 87-121.
[4] Kam-Fung Cheung, K.F., Bell, M.G.H.,
Bhattacharjya, J., Cybersecurity in logistics
and supply chain management: An overview
and future research directions, Transportation
Research Part E: Logistics and
Transportation Review, Vol. 146, February
2021, 102217.
[5] Raj, A., Mukherjee, A.A., Jabbour, A.B.L.S.,
Srivastava, S.K., Supply chain management
during and post-COVID-19 pandemic:
Mitigation strategies and practical lessons
learned, Journal of Business Research, Vol.
142, March 2022, Pages 1125-1139
[6] Rajesh, R., Pugazhendhi, S., Ganesh, K.,
Muralidharan, C., Sathiamoorthy, R.
Influence of 3PL service offerings on client
performance in India, Transportation
Research Part E: Logistics and
Transportation Review, Vol. 47, Issue
2, March 2011, Pages 149-165
[7] Lehtinen, J., Bask, A.H., Analysis of
business models for potential 3Mode transport
corridor, Journal of Transport Geography,
Vol. 22, May 2012, Pages 96-108
[8] Gunasekaran, A., Ngai, E.W.T., Information
systems in supply chain integration and
management, European Journal of
Operational Research, Vol. 159, Issue 2, 1
December 2004, Pages 269-295.
[9] Kayali, S., Turgay, S., Predictive Analytics
for Stock and Demand Balance Using Deep
Q-Learning Algorithm. Data and Knowledge
Engineering, (2023) Vol. 1: 1-10.
[10] Ashraf,M.H., Yuwen, C., Yalcin, M.G.,
Minding Braess Paradox amid third-party
logistics hub capacity expansion triggered by
demand surge, International Journal of
Production Economics, Vol. 248, June 2022,
108454
[11] Turgay, S., Güneş, A., Torkul, Y.E., Torkul,
B., Designing and Implementing Effective
Hospital Capacity Management Systems with
SEM Analysis. Social Medicine and Health
Management, (2023) Vol. 4: 37-46.
[12] Cheng, K.F., Bell, M.G.H., Bhattacharjya, J.,
Cybersecurity in logistics and supply chain
management: An overview and future
research directions, Transportation Research
Part E: Logistics and Transportation Review,
Volume 146, February 2021, 102217.
[13] Amiri, A.M., Ferguson, M.R., Razavi, S.,
Adoption patterns of autonomous
technologies in Logistics: evidence for
Niagara Region, Transportation Letters,
Volume 14, Issue 7, September 2022, Pages
685-696.
[14] Varriale, V., Cammarano, A., Michelino, F.,
Caputo, M., Integrating blockchain, RFID and
IoT within a cheese supply chain: A cost
analysis, Journal of Industrial Information
Integration, Volume 34, August 2023, 100486
[15] Casino, F., Dasaklis, T.K., Patsakis, C., A
systematic literature review of blockchain-
based applications: Current status,
classification and open issues, Telematics and
Informatics, Vol. 36, March 2019, pp. 55-81.
[16] İlter, İ., Turgay, S., Privacy Enhancement
with Perturbation Method for
Multidimensional Grid, Journal of Artificial
Intelligence Practice (2023), Vol. 6 Num. 4.
[17] Amiri, A.M., Ferguson, M.R., Razavi, S.,
Adoption patterns of autonomous
technologies in Logistics: evidence for
Niagara Region, Transportation Letters, Vol.
14, Issue 7, September 2022, pp.685-696
[18] Bhaktiar, P., Mursitama, T.N., So, I.G.,
Abdinagoro, S.B., Dwidienawati, D.,
Networking Capability and Learning
Capability as Determinants of Firm
Performance Mediated by Business Model
Innovation, WSEAS Transactions on
Information Science and Applications, vol. 20,
pp. 178-188, 2023,
https://doi.org/10.37394/23209.2023.20.21.
[19] Gregson, N., Work, labour and mobility:
opening up a dialogue between mobilities and
political economy through mobile work,
Mobilities, 18:6, 888-902, 2023.
[20] Agrawal, V., Goswami, P.K., Sarma, K.K.,
Week-ahead Forecasting of Household
Energy Consumption Using CNN and
Multivariate Data", WSEAS Transactions on
Computers, vol. 20, pp. 182-188, 2021,
https://doi.org/10.37394/23205.2021.20.19.
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
22
Volume 23, 2024
[21] Paridaens, H., Notteboom, T., Logistics
integration strategies in container shipping: A
multiple case-study on Maersk Line, MSC
and CMA CGM, Research in Transportation
Business & Management, Vol. 45, Part
C, December 2022, 100868.
[22] Raj, A., Mukherjee, A.A., Jabbour, A.B.L.S.,
Srivastava, S.K., Supply chain management
during and post-COVID-19 pandemic:
Mitigation strategies and practical lessons
learned, Journal of Business Research,
Volume 142, March 2022, Pages 1125-1139.
[23] Chatziioannou, I., Nikitas, A.,
G. Tzouras, P.G., Bakogiannis, E., Alvarez-
Icaza, L., Chias-Becerril, L., Karolemeas, C.,
Tsigdinos, S., Wallgren, P., Rexfelt, O.,
Ranking sustainable urban mobility indicators
and their matching transport policies to
support liveable city Futures: A MICMAC
approach, Transportation Research
Interdisciplinary Perspectives, Volume
18, March 2023, 100788.
[24] Kumar, S., Bhupati, R.R., Sharma, H.D.,
Misra, Y., Failure Log Analytics for Reducing
Electrical Machine Downtime using Deep
Learning", WSEAS Transactions on
Information Science and Applications, vol. 20,
pp. 444-452, 2023,
https://doi.org/10.37394/23209.2023.20.46.
[25] Doutoum, A.S., Eryigit, R., Tugrul, B.,
Classification of Guava Leaf Disease using
Deep Learning", WSEAS Transactions on
Information Science and Applications, vol. 20,
pp. 356-363, 2023,
https://doi.org/10.37394/23209.2023.20.38.
[26] Vlachos, I., Singh, R.K., Triad structure
impact on the triad resources-firm
performance relationship: Theory and
empirical evidence, Transportation Research
Part E: Logistics and Transportation Review,
Volume 179, November 2023, 103299
[27] Jianhong, W., Ricardo A. Ramirez-Mendoza,
R.A., Application of Interval Predictor Model
Into Model Predictive Control, WSEAS
Transactions on Systems, vol. 20, pp. 331-
343, 2021,
https://doi.org/10.37394/23202.2021.20.38.
[28] Siavash H. Sohrab, Some Implications of a
Scale Invariant Model of Statistical
Mechanics to Boltzmann versus Shannon
Entropy in Thermodynamics and Information
Theory, WSEAS Transactions on Computers,
vol. 20, pp. 56-65, 2021,
https://doi.org/10.37394/23205.2021.20.7.
[29] Xhabafti, M., Vika, B., Sinaj, V., A
Comparative Study of Statistical and Deep
Learning Model-Base weather Prediction in
Albania", WSEAS Transactions on Computer
Research, vol. 12, pp. 151-160, 2024,
https://doi.org/10.37394/232018.2024.12.15.
[30] Yang, D., Notteboom, T., Zhou, X., Spatial,
temporal and institutional characteristics of
entry strategies in inland container terminals:
A comparison between Yangtze River and
Rhine River, Journal of Transport
Geography, Vol. 90, January 2021, 102928
[31] Selviaridis, K., Spring, M., The dynamics of
business service exchanges: Insights from
logistics outsourcing, Journal of Purchasing
and Supply Management, Vol. 16, Issue
3, September 2010, pp. 171-184
[32] Sun, Y., Kirtonia, S., Chen, Z.L., A survey of
finished vehicle distribution and related
problems from an optimization perspective,
Transportation Research Part E: Logistics
and Transportation Review, Volume
149, May 2021, 102302.
[33] Shubailat, O., Al-Zaqeba, M., Madi, A &
Ababneh, A. (2024). Customs intelligence and
risk management in sustainable supply chain
for general customs department logistics.
Uncertain Supply Chain Management, 12(1),
387-398.
[34] Li, L., Assessing the relational benefits of
logistics services perceived by manufacturers
in supply chain, International Journal of
Production Economics, Vol. 132, Issue 1, July
2011, Pages 58-67
Contribution of Individual Authors to the
Creation of a Scientific Article (Ghostwriting
Policy)
- S.Turgay – investigation,
- Y.E.Yetiş, B.Erdemir- validation and
- Turgay writing & editing.
Sources of Funding for Research Presented in a
Scientific Article or Scientific Article Itself
No funding was received for conducting this study.
Conflict of Interest
The authors have no conflicts of interest to declare.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
WSEAS TRANSACTIONS on COMPUTERS
DOI: 10.37394/23205.2024.23.2
Yunus Emre Yeti
ş, Safiye Turgay, Bi
lal Erdemi
r
E-ISSN: 2224-2872
23
Volume 23, 2024
