The American Journal of Engineering and Technology
21
https://www.theamericanjournals.com/index.php/tajet
TYPE
Original Research
PAGE NO.
21-34
10.37547/tajet/Volume07Issue03-03
OPEN ACCESS
SUBMITED
01 January 2025
ACCEPTED
02 February 2025
PUBLISHED
03 March 2025
VOLUME
Vol.07 Issue03 2025
CITATION
Md Zahidur Rahman Farazi. (2025). Building Agile Supply Chains with
Supply Chain 4.0: A Data-Driven Approach to Risk Management. The
American Journal of Engineering and Technology, 7(03), 21
–
34.
https://doi.org/10.37547/tajet/Volume07Issue03-03
COPYRIGHT
© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.
Building Agile Supply
Chains with Supply Chain
4.0: A Data-Driven
Approach to Risk
Management
Md Zahidur Rahman Farazi
Department of Information Systems and Operations Management, The
University of Texas at Arlington
Abstract:
The aim of this study is to advance multi-label
delivery delay predictions in supply chains using
machine learning and deep learning models. The work
used Decision Trees, Random Forests, CNN, and FNN on
a real-life logistics dataset consisting of customers and
products features. EDA and feature selection were
examined and performed as a part of the data
preprocessing process at the pre-processing step of the
models. According to current model results, Random
Forest model reached maximum accuracy of 66.5%
along with Decision Trees and FNN. CNN, although,
worked well in some instances was not up to par in some
areas because it overfitted. The results also reveal how
Random Forest is a particularly useful algorithm for
predicting delivery delays accurately. The conclusion
suggests enhancing the deep learning models
performance and combining approaches. Further work
should also incorporate other variables in order to
improve the predictive capability in real-life
requirements of supply chain environments including
conditions and stocks.
Keywords:
Supply Chain 4.0, Machine Learning, Deep
Learning, Risk Management.
Introduction:
Supply Chain 4.0 is fully embedded
internet of thing, analytics, automation and data- driven
decision making in order to manage the supply chain
process. These core processes integration of DL and ML
helps organizations process large amounts of real-time
data for predictive insights, risk reduction and proactive
response to disturbances [1]. These technologies aid in
The American Journal of Engineering and Technology
22
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
identifying possible threats because, based on
historical data fed into the systems, the algorithms get
to adjust their projections continually as the new data
stream in. As a result, companies can have full chain
visibility,
efficient
resource
managing
and
enhancement their capacity to respond to risks and
volatilities [2].
The use of data analytics in the approaches is a
revolutionary step from managing supply chain risks in
a reactive-fashion to managing them in an
anticipatory-fashion [3]. From this perspective, DL and
ML enable organizations to improve their performance
to a higher level, optimizing not only essential business
processes but also providing the prospects for complex
organizational growth in today’s environment when
flexible response and fast decision-making becomes
critical [4].
The increasing dynamism and sophistication of
supplies means that the ability to predict delivery delay
has become a huge problem as such delays may be
influenced by spatial issues, type of products, or
customers [5]. Most supply chain practitioners are
usually tasked with predicting more than one form of
delay at one time in a supply chain which makes it a
multi-label prediction problem [6]. Such imprecise risk
management leads to higher operational costs,
resource consumption, and customer dissatisfaction
that requires developing enhanced methods for
managing delivery risks as efficiently as possible.
The proposed study therefore seeks to design and
apply complex ML and DL algorithms to increase the
reliability of multi-label delivery delay prognosis in
value chains and assist organizations in reducing risks
and improving operational performance. The
objectives of the research are:
•
To determine and categorize delivery delays
factors in supply chains.
•
To perform machine learning and deep
learning algorithms in order to compare the results of
multi-label delivery delays prediction.
•
To test the efficacy of such models in real
conditions for risk detection and operational decision
makings.
•
To offer practical suggestions regarding the
application of these models into the existing supply
chain theories for improving supply chain adaptability
and flexibility.
Literature Review
A.
Understanding of the Fourth Industrial
Revolution and Supply Chain 4.0
Machine learning (ML) and neural networks (NN) have
Industry 4.0 has brought outdated innovative
technological changes that have affected the supply
chain management of numerous industries. Supply
Chain 4.0 which is the focal piece of Industry 4.0
employs IoT, AI, ML, and DL to improve organizational
supply chain effectiveness, real-time tracking, and
flexibility [7]. In this literature review, the definition,
development, the components and the significance of
agility and risk management as fundamental to the
current supply chain concepts are discussed.
Supply Chain 4.0 is a deviation from the typical supply
chains that exhibited limited integration and excessive
complexity. These conventional models could not meet
the demand for flexibility that is necessary for engaging
with changing customer needs [8]. Supply Chain 4.0 can
be originated from the early internet and the enterprise
resource planning (ERP) system that coordinated the
supply chain activities and information flows. This has
however been propelled by IoT, AI, ML and data
analytics that have shaped inter-connected physical and
digital systems for real-time data generation and
decisions [9].
The American Journal of Engineering and Technology
23
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
Figure 1: Supply Chain Enablers
(Source: [14])
Several technologies are depicted in the diagram that
support Supply Chain 4.0 such as; Sensors, Robotics
and Automation, Big Data, Cloud services, 3DPrint and
RFID which increases real time monitoring, demand
forecasting and new type of production [14].
Supply Chain 4.0 depends on several important
technologies, some of the ways that IoT will impact the
supply chain include collecting and transmitting data
about inventory, assets and production that occurs in
real time from various points on the supply chain [10].
AI entail large datasets for pattern recognition, making
decisions and predicting future situations hence, useful
trade activities such as demand forecast and risk
managing [11]. Automated decision-making can
enhance organizational operation since ML, a subfield
of AI, increases machines’ performance by training on
data; its application includes quality control and
prediction of equipment failures [12]. DL is a further
development of the ML system that uses neural
networks; this algorithm works better when solving
such issues as image identification or detecting
unusual activity. Robots’ technology and automated
guided vehicles, lessen the labor cost and improve the
productivity of the supply chain operation; and
advanced analytics, particularly the prognostic and
diagnostic ones, help in risk assessment and decision
making for supply chain management [13].
B.
Flexibility and risk are the major components
of Supply Chain 4.0
Agility on
the other hand, is the supply chain’s capacity
to rapidly respond to disruption or demand volatility
and risk management centers around recognizing,
evaluating, and managing risks [15]. These factors play
specific and significant roles in Supply Chain 4.0,
according to several researchers. In the paper [16]
discussed agile supply chains can provide more
response to disruptions and help organizations satisfy
the customer demand in volatile contexts. Furthermore,
the study [17] shows that the management strategies
are useful for managing supply chain risks and
increasing performance, thus stabilizing operations.
As a result, it is correct to mention that Supply Chain 4.0
outlines a new approach to manage supply chains.
Incorporating smart technologies in an organization
provides organization with a boost in efficiency gain,
visibility and flexibility. However, for Supply Chain 4.0 to
be fully operational to provide all the benefits that come
with it, there is need for agility, plus wiser ways of
performing risk management due to the dynamic
nature. Looking to the future of Supply Chain 4.0, it can
be expected that as the technologies get improved and
developed, Supply Chain 4.0 will help to stimulate
further
productivity
change
in
supply
chain
management.
C.
Problems Associate with Conventional Supply
Chain Management
Traditional supply chain management has a number of
important challenges that prevent supply chain to
function optimally and effectively, which is explained by
low level of structure flexibility and high percentage of
The American Journal of Engineering and Technology
24
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
manual processes [18]. On the same account, one
major weakness of traditional supply chain is lack of
flexibility; it is relatively difficult for a supply chain that
is rooted in traditional model to make drastic changes
when change is inevitable due to its negative impacts
that it will bring about such as increased costs and
reduced customer satisfaction [19]. The research [20]
discussed that the increased demand may lead to
higher stockouts and rationing due to insufficient
production capacity or fixed supply arrangements.
Also, a lack of transparency is rife in normal supply
chains, which hinders an organization’s capacity to
track the flow of the products and supplies. This lack of
insight can lead to inefficiencies, delays and costs
cutting; for example, concerns such as delayed or lost
shipments may only discovered after the product has
arrived with the customer, resulting into customer
dissatisfaction and monetary loss. In addition, the
conventional supply chain is relatively rigid in managing
changing dynamics such as natural disasters or an
economic recession that has a huge cost implication and
negative impact to the organization’s reputation [21].
For example, the natural disaster that shuts down much
of manufacturing such as a factory closes down supply
chain which slows down productions.
Figure 2: Conventional Supply Chain Management
Source: [21]
Further adding to these difficulties are various risks
occurring to conventional supply chains such as
demand volatility, geo political risks, and disasters.
Fluctuations in consumer demand create undesirable
conditions such as stock out conditions and conversely
inventory conditions, political instabilities trade wars
cause disruptions in the right of transport of goods.
Hurricanes or any other natural disasters increase
these problems by destroying infrastructure, slowing
transport, and failing to supply sufficient products [22].
Taken together, the drawbacks and dangers of supply
chain management as practiced in the traditional
system require rethinking and improvement work.
The conventional approaches of risk assessments and
delay predictions call for the use of history and
statistics; not enough provide the outstanding
forecasts needed to handle emerging risks [23]. These
populations may also take a long time to be awakened
to respond to incidents hence can be vulnerable and
be able to lose a lot of money. In an attempt to address
these problems organizations are gradually beginning
to seek better analytical technologies and approaches
to supply chain management. Such are the adoption of
artificial intelligence, machine learning, as well as use
of data in increasing visibility and optimizing outcomes
and minimizing risks. Through adoption of these
technologies, organizations can establish strong,
effective and adaptive supply chains capable to
managing itself within today’s environment.
D.
The Use of Artificial Intelligence in Supply Chain
4.0
Application of AI can help organizations to improve
decision-making by offering automation, optimization
and increased control of organizational data, relying on
large data sets [24]. The study [25] demonstrated that
AI is instrumental in enabling primary supply chain tasks
including demand forecasting, route management, and
predictive maintenance. All of these applications help in
achieving cost reduction efforts, enhancing service
delivery, and managing risks occasioned by disruption of
supply chains.
Another field of supply chain management, demand
forecasting, has relatively clear and potentially
enormous benefits from AI models for predictions [26].
Unlike the former tools that have been used in
estimating demand which can be done by using
historical data and where analysis is again performed
manually, was not able to capture changes of real time
markets. Machine learning models, a type of on AI
algorithms outperforms human derived models in this
aspect because the data that feed into the model can in
The American Journal of Engineering and Technology
25
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
real time data such as the prevailing weather, the state
of the economy and society at that particular time [27].
The study [28] showed that the use of AI increases the
accuracy of demand forecasting by 30% if compared to
the traditional approaches. Similarly, the study [29]
pointed to corresponding advantages of applying AI in
stock cost optimization by improving demands’
prognosis.
Another process area that has benefited from use of AI
technology is in route optimization. Machine learning
is then employed to forecast the precise delivery
routes which with respect to congestion, fuels and
delivery time [30]. In the research [31] reported that
the use of AI for route planning and optimization
minimized delivery time by 15% and had a
corresponding impact on the cost of logistics by
minimizing it by 20%. This is a major enhancement
compared
with
traditional
route
optimization
approaches that have dynamic problem-solving
difficulties, such as in situations that involve changes in
roads or weather conditions.
Another strategic application of AI in Supply Chain 4.0 is
predictive maintenance, as it highlights the value of
minimizing downtime and increasing the reliability of
equipment [32]. Other maintenance schedules are
based on time hence resulting to either over
maintenance or under maintenance of the assets.
Conventional concept of condition monitoring on the
other hand is based on causative failures and typically
involves the use of sensor data to determine when an
asset is likely to fail next, thus shorting the operation
time [33].
Figure 3: AI-Based Predictive Maintenance
Source: [34]
The study [34] showed that AI-based predictive
maintenance decreased equipment failure rate in
contrast to the reactive maintenance. However, some
challenges still persist. Despite the AI’s advances in
supply chains, challenges such as data quality and
integration hinder AI effectiveness, are common
concerns surrounding implementation among many
organizations. Moreover, while previous works
defined the promising applications of AI, less recent
research is applying increasingly efforts on practical
issues of putting such applications into effect, including
for instance scalability and ethics.
E.
Applying Machine and Deep Learning for
Supply Chain Risk Management
The machine learning (ML) and deep learning (DL) have
grown to be high-impact solutions for handling the
problems the traditional supply chains. These
technologies provide enhanced features of predictive
statistics, risks assessment, and optimization that help
various organizations to reduce risks and make better
decisions.
Some of the most common machine learning models
used in supply chain risk management include Random
Forest, Support Vector Machines (SVM), and XGBoost.
The generated models can be applied to anticipate the
lead time, estimate risks on the supply chain and
improve on its working [35]. For instance, Random
Forest can be applied to forecast demand variation and
SVM
—
to conduct anomaly detection in the supply
chain data.
There are many works that have shown that using ML in
supply chain planning and decision-making is beneficial.
The study [36] showed that by adopting ML models,
demand can be more accurately forecasted than by
traditional forecasting methods across many products.
Although the study [37] found that consequent ML
algorithms can be used in helping to predict potential
risks in the SCM, which may range from transportation
hitches to supplier bankruptcy.
DL which is a subfield of ML is also being used more in
supply chain analysis. DL models such as neural
networks, CNNs, and RNNs are capable of recognizing
intricate patterns and dependencies in patterns, which
is appropriate for tasks like multi-label classification
[38]. In supply chain management, DL models can be
employed to make forecast on delivery delays
depending on the weather condition, mode of transport
used and performance of suppliers.
Sometimes DL approaches outperform the traditional
ML models where there are complex pattern and large
dataset. In the research [39] observed that using DL
models enhance prediction of delivery delays in a large
e-commerce firm as compared to ML models. However,
its known that DL models can be computationally
The American Journal of Engineering and Technology
26
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
expensive in their training process and may even need
GPUs [40].
Therefore, it can be seen that ML and DL contains
useful tools useful for handling the issues affecting the
existing supply chain. Through the use of the
mentioned technologies, it is possible to enhance the
capacity of organizations to try and estimate potential
risks, increase operational efficiency, and make
rational decisions. That is why as more and more of
these technologies develop, in the research can be
sure that even more applications in the field of supply
chain risk management will be developed.
F.
Multiple label classification issues in supply
chain management
The problem of handling multi-label classification
scenarios in SCM is challenging because several labels
are correlated and interdependent in supply chain
management risk and return scenarios. Multi-label
classification is different from single label classification
dealing with cases in which multiple outcomes such as
transport delay, stock-out, and supplier disruption are
possible [41]. These complexities make multi-label
classification very important in handling real time
decisions within the supply chain. The paper [42] imply
that multiple-label classification is significant when
anticipating disruptions in the supply chain but claim
that numerous dangers are not efficiently explained by
the current frameworks.
Consequently, the paper [43] deal with multi-label
classification, binary relevance and one vs rest which
enhanced the accuracy in the supply chain delay
prediction. Nevertheless, addressing the issues related
with the imbalanced data is always a concern. This
problem has, however, been dealt with fairly well by
the application of SMOTE (Synthetic Minority Over-
sampling Technique) which was endorsed in the study
by [44] for the development of SMOTE based multi-
label models for delay prediction.
Predictive analytics has also added another promising
perspective to risk reduction in Supply Chain 4.0.
Whereas previous strategies and future forecasting
have been utilized in previous literature, the study [45]
explained that real-time and predictive models-
oriented concepts can effectively enhance supply
chain visibility. Other approaches related to the use of
artificial intelligence can also be recommended
because the fundamental elements of AI-based
techniques have been tested experimentally in other
cases and effectiveness was proven in terms of the
ability to adapt to changing uncertainties more quickly
than traditional means.
G.
Gaps in Existing Literature
The current research focusing on delivery delay
prediction in supply chains presents several limitations,
concerning, for example, the application of multi-label
predictive models and their application for real-time risk
assessment. Despite various ML ad DL applications in
resolute supply chain functions, many of these works
employ single-label classification approaches for delay
prediction. Validation of traditional approaches come
with various assumptions that disregard how real-world
supply chain are complex and could be subject to
various delays [46]. The concept of the multi-label
predictive models that enable the practice of the
numerous outcomes simultaneously is relatively
underdeveloped; hence, the applicability of this concept
in dynamic world is less.
Furthermore, while the application of AI technologies
can impr
ove supply chain’s flexibility and robustness,
most of the research works are based on the analysis of
archive data rather than monitoring risks in real-time
manner. The research [47] highlighted the importance
of using predictive analytics for supply chain
management while admitting that the integration of
real-time decision-making models may a challenging
task because of the data quality and integration
problems. Today’s scholarly work does not provide an
extensive framework for integrating multi-label models
into real-time decision-making systems.
The future work should be oriented to utilize DL and ML
to develop high-impact, fast-action, accurate multi-label
predictive models for solving multiple risks at a time for
enhancing the availability and agility of the supply chain.
ML and DL were described in the literature as the keys
to developing new and highly adaptive and re-
illuminated supply chain management structures with
high risk-management features. Although currently not
as actively investigated as single-label approaches,
multi-label classification methods encompassed more
accurate solutions to supply chain difficulties like
delivery delay and disruptions. Researches focused on
how accurate predictive analytics were and how real-
time decision making and response could be leveraged
by AI models. Future research was suggested to
concentrate on enhancing multi-label approaches and
incorporating it within elaborate supply chain models.
Companies were urged to harness sophisticated AI tools
and approaches
to improve the firm’s adaptability and
reduce perils of operating in dynamic contexts.
METHODOLOGY
This work utilizes comparisons between a machine
learning (ML) and a deep learning (DL) based models for
multi-label delivery delay prediction. The basic
variables, which cause delay are determined by data
gathering and cleaning the data. To assess model
The American Journal of Engineering and Technology
27
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
performance, actual, real-time datasets are used to
identify the risk and the proposal brings enhanced
decision making to the supply chain.
A.
Data Collection
This data set for this research work was obtained from
Kaggle and consist of actual logistics and supply chain
data. It yielded 15,549 records and 41 variables which
provide a broad perspective of numerous aspects that
affect supply chain. They are, payment in respect of
type of payment which indicates the methods used in
the transactions, and profit per order which indicates
the profitability every order. Furthermore, the dataset
contains the daily sales per customer, which gives
information about customer-oriented sales, beside the
category ID and the category name, which categorizes
products. Other attributes include geographical and
identification indicators, including customer city,
country, and ID number, and customer segment
information that distinguishes between customers
according to their intent and age. This high quality and
multi-faceted data are ideal for performing a
comprehensive analysis across all delivery delay
categories and to model how specific factors play a role
in them [48].
B.
Proposed Architecture
The following architecture has been proposed for the
prediction of multiple labels of delivery delays:
Machine learning ML: Decision tree and Random
Forest Deep learning DL: Convolutional neural
networks CNN and feedforward neural networks FNN.
Every model that has been chosen in this paper has
been done so because of its merit as found in prior
research and applicability in solving the multi-label
prediction problem in the supply chain. Table 3
presents each model and how it can be justified with
findings from previous studies; the reason behind
selecting each model.
1)
Decision Tree
Decision Trees can be found extensively in various
fields including supply chain management and
predictive
modeling
because
of
their
easy
interpretability [49]. The study [50] reported that DT
are useful in supply chain risk management
considering their ability to provide decision-makers
with clear decision-making trees and risk factor
information. In single-label classification tasks, the
study [51] also reported Decision Trees effective for
predicting supply chain disruptions. In this research the
Decision Tree model stands out to set the benchmark
given its interpretability. The hierarchical structure
also helps locate priority factors about delays in
delivery [geographical location, type of products].
While its performance decreases in multiple class
multiple label scenario it has the advantage of being
able to give insights into the importance of features
during the initial stages of the detection process.
2)
Random Forest
Random Forest is another technique that integrates
multiple decision trees which are more effective than
the performance of an individual tree according to the
outcomes of numerous research [52]. The research [53]
proposed Random Forest, which means that it has good
generalization performance since it randomly average
out the decision trees. This research selects Random
Forest for its ability to deal with large numbers of
features in the dataset. Random Forest classifies the
delivery delays based on the combined decision paths
which capture complex variable interdependencies such
as customer segments with product categories, thus
improving the model’s overall prediction.
3)
Convolutional Neural Network (CNN)
The most popular type of neural networks, CNNs has
been used mainly for image recognition but recently the
use of CNNs for structured data is on the rise [54]. Study
[55] have pointed out that CNNs can also be used in text
classification tasks pointing out that such networks are
capable of learning patterns in structured data other
than images. Moreover, study [56] applied the CNNs in
demand forecasting in supply chain and their results
reveal the enhanced accuracies in the forecasting due to
the capitalization of the relations both in space and
time. CNNs are chosen for this study because participant
CNNs can detect local relationships between customers’
demographics, product offering, and delivery locations.
The capability of CNNs to process the grid-like inputs
data make them useful in revealing hidden correlations
making multi-label classification more accurate. In this
paper, CNNs will be extended to handle highly
structured supply chain data and its finer characteristics
that motivate delivery delays. Through the use of such
convolutional layers the model will able to detect
transformative interactions of the variables for resulting
in multi-label delay with better prediction capacity.
4)
Feedforward Neural Network (FNN)
Feedforward Neural Networks (FNNs) are the simplest
design of the Deep Learning model for structured data
[57]. The study [58] confirmed FNNs’ presentation of
day-to-day non-linear interaction between variables
which make them suitable for supply chain forecasting
and optimization. FNNs were selected for this study
because they perform well in relation to handling of
complex relations in multi-label prediction activities.
FNNs on the other hand are trained using big data and
it extracts vital patterns which are necessary for
accurate predictions unlike other Machine Learning
models. The FNNs will be used in this study to capture
The American Journal of Engineering and Technology
28
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
such complexities between features like the number of
sales per customer, customer’s city, and their product
preferences since FNNs’ multi
-layer architecture
powers the modeling of non-linear complexities of
features for better multi-label delay predictions.
C.
Data Analysis
The data analysis is a critical stage in the process of raw
supply chain data and generating an outlook for
delivery delays. For this study, the dataset contains
demographical data about customers as well as
regarding the products like customer city, country,
segment among others: category name, profit per
order, payment type sales per customer, among
others. All affect delivery outcomes and therefore
required a careful examination of the distributions and
interactions with the target variable-delivery delays.
Then, the feature selection is performed based on the
EDA in order to gather more information on the
structure and distribution of the dataset [59]. The
distributions of key features such as location and order
profitability are done using descriptive statistics to look
out for patterns or outliers that may result to delays.
For instance, based on the visualization of the data set
for the correlation between the customer country and
delivery delays, the areas that are most affected by
delivery delays for reasons such as logistics constrains
and regulation of customs may be identified.
Feature selection techniques are then used in an
attempt to determine which of these variables have
significant influences to the delay variable. Most feature
selection methods such as Correlation analysis and
feature importance rankings from models like Random
Forest provide an understanding of how much
predictive power a feature has; -therefore, it can help in
the dimensionality problems allowing for better
accuracy in the model. Customer city and product
category are among the significant predictors of delay
that we anticipate in the analysis.
Data preprocessing comes next before removing rows
containing missing attributes, scaling numeric attributes
and encoding categorical features in order to feed the
machine learning models. Last but not least, key findings
are presented using heat maps and bar charts,
explaining further how features in the data set
interconnect and cause delivery delays. In conclusion,
the data analysis phase provides a strong foundation for
model development while also guiding the selection of
features and architecture.
Figure 4: Data Analysis Flow Chart
RESULTS AND DISCUSSION
This work sought to improve the accuracy of multi-
label delivery delay predictions in supply chains, using
enhanced machine learning and deep learning models.
This research used Decision Trees, Random Forests,
Convolutional
Neural
Networks
(CNN),
and
Feedforward Neural Networks (FNN) with data
containing other attributes of the customers and the
products needed to identify the causes of delivery
delays. The models were then checked of their
predictive accuracies and out of these the best method
was selected in order to effectively deal with various
real life like supply chain problems. The subsequent
section of the paper outlines the analysis results,
studies’ findings and insights.
The American Journal of Engineering and Technology
29
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
Figure 5: Distribution of Profit Per Order
From the fig 5, it is observed that the histogram
representing profit per order has inflated mean with
many orders being placed near to zero. This means that
the majority of the orders are either marginally
profitable or actually losing money, with a few
extremes of large losses at the far left. This implies that
there are many orders that are nearly profitable and
some orders that may be unprofitable more analysis
could be done to check profitability.
Figure 6: Avg. Sale per Costumer by Category
The fig 6 illustrates how sales distribution is in relation
to different product categories. That is a major
observation that points to a higher average in
‘Consumer Electronics,”” indicating that it is the most
productive category. As could see in the value of ‘Sales
per Customer’ other categories such as ‘Fitness
Equipment,’ ‘Golf Apparel,’ are havin
g high values
revealing its comparison with other categories such as
‘Accessories,’ ‘Baseball,’ which have low values of ‘Sales
per Customer.’ From this, we deduce that the customer
expenditure distribution is skewed towards a few
important groups.
Figure 7: Correlation Heatmap
The American Journal of Engineering and Technology
30
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
The fig 7 helps to understand how features of the
dataset are related to each other. Gray colors mean
negative correlation, black and white mean no
correlation, and lighter spherical colors mean positive
correlation. For example, the qualitative variable
profit_per_order is positively and significantly related
to both order_item_profit_ratio as well as sales which
suggest that selling a greater number of orders and
higher profit margins directly results in higher profits
per
order.
Also,
features
such
as
the
order_item_total_amount and sales, present positive
significant correlation, thus indicating the order value as
influenced by these features directly. Longitude and
the order_item_quantity variables are least associated
with most of the factors, meaning that they are least
relevant with predictive predisposition.
A.
Comparison
Figure 8: Models Acracy Compression
The fig 8 displayed the accuracy percentages of four
machine learning models: These include Decision Tree,
Random Forest, Feedforward Neural Network, and
Convolutional Neural Network. The average of
accuracy is calculated to be 66.50% when it comes to
Random Forest model, the model that has a high
degree of accuracy in its predictions. Next is the
Decision Tree whose accuracy stands at 62.57% a
relatively decent but could be better position. The
Feedforward Neural Network bears a lower accuracy of
60.10 % here which shows the issues of the network to
predict accurately. The Convolutional Neural Network
has the lowest accuracy at 57.56% which also shows
that generalization is a problem in this case and further
work should be done in improving model training and
optimization.
Table 1: Matrices Comparison
In the evaluation of machine learning models for
predictive analysis, four models were compared: The
classification models include Decision Tree, Random
Forest,
Feedforward
Neural
Network
(FNN),
Convolutional Neural Network (CNN). The results
reveal distinct performance characteristics among
these models based on key metrics: Precision, Recall,
F1 score, and Accuracy is what other languages use
while analyzing the results of a model.
Through this experiment, the Random Forest model
claimed the highest accuracy at an acceptable value of
66.5% from all the methods tested. Through this, it
obtained a precision of 70.0 % and a recall of 73.0 %
which means it is precise in identifying the positive cases
but could still improve on precision. Compared to other
algorithms Decision Tree had an accuracy of 62.6% in V2
Model
Accuracy
(%)
Precision
(%)
Recall (%)
F1 Score
(%)
Decision Tree
62.572347
67.02938
68.826816
67.916207
Random Forest
66.495177
70.042872
73.01676
71.498906
Feedforward
Neural
Network
60.096463
64.334204
68.826816
66.504723
Convolutional
Neural
Network
57.55627
57.55627
100
73.061224
The American Journal of Engineering and Technology
31
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
and bad result in recall 68.8 and it proved to be highly
susceptible to over fitting in case of highly complex
datasets.
The Feedforward Neural Network on the other hand
had the overall accuracy of 60.1% which though
equalized the recall of the Decision Tree. The CNN
model had the highest recall of 100 % but the accuracy
and precision which were 57.6% proved that the model
tend to overfit the positive class. In summary, it might
be said that Random Forest is characterized by the
highest level of balance, though all the models
introduced in this paper require further improvements
to work as primary tools for enhanced prediction.
DISCUSSION
The results of this study support the claims of this
study and determine that the Random Forest is the
most accurate model for estimating delivery delay in
supply chain at a rate of 66.5%. This is supported by
prior research that has demonstrated Random Forest’s
resilience in processing high numbered, many-
dimensional variables (Breiman, 2001; Mishra et al.,
2020). The Decision Tree model, though easy to
interpret, has low accuracy (62.6%) and recall as
mentioned by Mardani et al. (2017) that the model
overfits the data in enlarged datasets. In addition, the
Feedforward Neural Network (60.1 percent) had clues
to the efficiency of identifying extensive patterns but
did not exercise proficiency in extracting complex
patterns than the Random Forest, as discovered by
Yildirim et al., (2019). Whereas the Convolutional
Neural Network achieved a high recall of 100%, its low
accuracy of 57.6% indicated overfitting problems,
which have been pointed by Kim (2014) as a potential
drawback of using the Convolutional Neural Network
for non-image data. In general, the Random Forest
model is as seen above the best predictor; however,
future studies should focus on improving the
performance of the other models.
CONCLUSION
To summarize, this study focused on improving the
multi-label delivery delay predictions in the context of
supply chain by using Decision Trees, Random forests
CNNs and FNNs. This showed that Random Forest
model had the highest accuracy in compare to other
models used in the analysis in terms of predictive tasks.
Still, CNNs faced issues like overfitting while FNNs
offered only moderate accuracy in case of the analyzed
models, but all three provided meaningful information
about potential delivery delay in light of certain
attributes of customer and products.
The future work should therefore be directed towards
streamlining of these models and examination of
various optimization procedures that will enhance the
predictive capability of such models in real-life chain
environment. To make the predictions more reliable,
other data related to real time traffic, atmospheric
conditions during the course of the day, and stock
conditions can be integrated into the system. Further,
research should be conducted further to derive adaptive
products that combine some theoretical approaches
with other models which could be much more effective.
Practice implications are to adapt the Random Forest
model in their functioning and provide delay predictions
with higher accuracy, using data for model updates.
Fresh models will also need to be put in place
periodically as changes are likely to arise from time to
time within supply chain environments. In sum, the
study contributes to the literature by showing that
specific risk factors need to be mitigated through
applications of advanced analytics in SCM for improved
performance.
REFERENCES
E. Manav
alan and K. Jayakrishna, “A review of Internet
of Things (IoT) embedded sustainable supply chain for
industry 4.0 requirements,” Computers & Industrial
Engineering, vol. 127, no. 1, pp. 925
–
953, Jan. 2019, doi:
https://doi.org/10.1016/j.cie.2018.11.030.
A.
Spieske and H. Birkel, “Improving supply chain
resilience through industry 4.0: a systematic literature
review under the impressions of the COVID-19
pandemic,” Computers & Industrial Engineering, vol.
158,
p.
107452,
Jun.
2021,
doi:
https://doi.org/10.1016/j.cie.2021.107452.
M. Pavan and L. Samant, “Digitalization and E
-
Commerce Trends in the Industry,” pp. 253–
272, Jan.
2024, doi: https://doi.org/10.1007/978-981-97-6577-
5_13.
Z. Latinovic and S. C. Chatterjee, “Achieving the promise
of AI and ML in delivering economic and relational
customer value in B2B,” Journal of Business Research,
vol.
144,
pp.
966
–
974,
May
2022,
doi:
https://doi.org/10.1016/j.jbusres.2022.01.052.
Z. Zong, T. Feng, T. Xia, D. Jin, and Y. Li, “Deep
Reinforcement Learning for Demand Driven Services in
Logistics and Transportation Systems: A Survey,”
arXiv.org,
Mar.
23,
2022.
https://arxiv.org/abs/2108.04462
H. N. Perera, J. Hurley, B. Fahimnia, and M. Reisi, “The
human factor in supply chain forecasting: A systematic
review,” European Jo
urnal of Operational Research, vol.
274, no. 2, pp. 574
–
600, Apr. 2019, doi:
https://doi.org/10.1016/j.ejor.2018.10.028.
B. Unhelkar, S. Joshi, M. Sharma, S. Prakash, A. K. Mani,
and M. Prasad, “Enhancing supply chain performance
using RFID technology and decision support systems in
The American Journal of Engineering and Technology
32
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
the industry 4.0
–A systematic literature review,”
International Journal of Information Management
Data
Insights,
vol.
2,
no.
2,
2022,
doi:
https://doi.org/10.1016/j.jjimei.2022.100084.
M. V. Chester and B. Allenby, “Toward adapt
ive
infrastructure: flexibility and agility in a non-
stationarity
age,”
Sustainable
and
Resilient
Infrastructure, vol. 4, no. 4, pp. 1
–
19, Jan. 2018, doi:
https://doi.org/10.1080/23789689.2017.1416846.
J. Sutduean, A. Singsa, T. Sriyakul, and K.
Jermsittip
arsert, “Supply Chain Integration, Enterprise
Resource Planning, and Organizational Performance:
The Enterprise Resource Planning Implementation
Approach,” Journal of Computational and Theoretical
Nanoscience, vol. 16, no. 7, pp. 2975
–
2981, Jul. 2019,
doi: https://doi.org/10.1166/jctn.2019.8204.
T. de Vass, H. Shee, and S. Miah, “IoT in Supply Chain
Management: Opportunities and Challenges for
Businesses in Early Industry 4.0 Context,” Operations
and Supply Chain Management: An International
Journal, vol. 14, no. 2, pp. 148
–
161, Jan. 2021, doi:
https://doi.org/10.31387/oscm0450293.
Farazi, M. Z. R. (2024). “Evaluating the impact of AI and
blockchain on credit risk mitigation: A predictive
analytic
approach
using
machine
learning.”
International Journal of Science and Research Archive,
13(1),
575
–
582.
https://doi.org/10.30574/ijsra.2024.13.1.1707
Y. R. Shrestha, V. Krishna, and G. von Krogh,
“Augmenting organizational decision
-making with
deep learning algorithms: Principles, promises, and
challenges,” Journal
of Business Research, vol. 123, pp.
588
–
603,
Feb.
2021,
doi:
https://doi.org/10.1016/j.jbusres.2020.09.068.
A. Bhargava, Mohd. Suhaib, and A. S. Singholi, “A
review of recent advances, techniques, and control
algorithms for automated guided vehicle system
s,”
Journal of the Brazilian Society of Mechanical Sciences
and Engineering, vol. 46, no. 7, Jun. 2024, doi:
https://doi.org/10.1007/s40430-024-04896-w.
Farazi, M. Z. R. (2024). “Designing circular supply
chains with digital technologies for competitive
sustainability:
An
operation
management
perspective.” International Journal of Science and
Research Archive, 13(1), 2346
–
2359.
https://doi.org/10.30574/ijsra.2024.13.1.1928.
R. K. Singh and S. Modgil, “Adapting to disruption: the
impact
of
agility,
absorptive
capacity
and
ambidexterity
on
supply
chain
resilience,”
International Journal of Productivity and Performance
Management,
Aug.
2024,
doi:
https://doi.org/10.1108/ijppm-01-2024-0057.
W. Ahmed and S. Huma, “Impact of lean and agile
strategies on supply cha
in risk management,” Total
Quality Management & Business Excellence, vol. 32, no.
1
–
2,
pp.
1
–
24,
Oct.
2018,
doi:
https://doi.org/10.1080/14783363.2018.1529558.
M. Nakano and A. K. W. Lau, “A systematic review on
supply chain risk management: using the strategy-
structure-process-
performance
framework,”
International Journal of Logistics Research and
Applications, vol. 23, no. 5, pp. 443
–
473, Dec. 2019, doi:
https://doi.org/10.1080/13675567.2019.1704707.
X. Wang, V. Kumar, A. Kumari, and E. Kuzmin, “Impact of
Digital Technology on Supply Chain Efficiency in
Manufacturing Industry,” Lecture Notes in Information
Systems and Organisation, vol. 54, pp. 347
–
371, 2022,
doi: https://doi.org/10.1007/978-3-030-94617-3_25.
T. J. Pettit, K. L. Croxton, and J. Fiksel, “Th
e Evolution of
Resilience
in
Supply
Chain
Management:
A
Retrospective on Ensuring Supply Chain Resilience,”
Journal of Business Logistics, vol. 40, no. 1, pp. 56
–
65,
Mar. 2019, doi: https://doi.org/10.1111/jbl.12202.
M. S. Sodhi, C. S. Tang, and E. T. Willenson, “Research
opportunities in preparing supply chains of essential
goods for future pandemics,” International Journal of
Production Research, vol. 61, no. 8, pp. 1
–
16, Feb. 2021,
doi: https://doi.org/10.1080/00207543.2021.1884310.
J. Jeng, S. J. Buckley, H. Chang, and J. Schiefer, “BPSM:
An Adaptive Platform for Managing Business Process
Solutions,”
ResearchGate,
Oct.
29,
2011.
https://www.researchgate.net/publication/251735269
_BPSM_An_Adaptive_Platform_for_Managing_Busines
s_Process_Solutions (accessed Oct. 02, 2024).
M. Kunkel, “Lessons from a Hurricane: Supply Chain
Resilience in a Disaster, An Analysis of the US Disaster
Response to Hurricane Maria,” conservancy.umn.edu,
Sep.
2020,
Available:
https://conservancy.umn.edu/items/a78b1577-9e0a-
4871-b562-94b8525ba1fc.
P. G. George and V. R. Renjith, “Evolution of Safety and
Security Risk Assessment methodologies to use of
Bayesian Networks in Process Industries,” Process
Safety and Environmental Protection, Mar. 2021, doi:
https://doi.org/10.1016/j.psep.2021.03.031.
N. K. Rajagopal et al., “Future of Business Culture: An
Artificial Intelligence-Driven Digital Framework for
Organization Decision-
Making Process,” Complexity,
vol. 2022, no. 2, pp. 1
–
14, Jul. 2022, Available:
https://www.hindawi.com/journals/complexity/2022/7
796507/
O. A. Elufioye, C. U. Ike, O. Odeyemi, F. O. Usman, and
N. Z. Mhlongo, “AI
-DRIVEN PREDICTIVE ANALYTICS IN
AGRICULTURAL SUPPLY CHAINS: A REVIEW: ASSESSING
The American Journal of Engineering and Technology
33
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
THE BENEFITS AND CHALLENGES OF AI IN
FORECASTING DEMAND AND OPTIMIZING SUPPLY IN
AGRICULTURE,” Computer Science & IT Research
Journal, vol. 5, no. 2, pp. 473
–
497, Feb. 2024, doi:
https://doi.org/10.51594/csitrj.v5i2.817.
Farazi, M. Z. R. (2024). “Exploring the Role of Artificial
Intelligence in Managing Emerging Risks: An In-Depth
Study of AI Applications in Financial Institutions’ Risk
Frameworks.” The American Journal of Management
and
Economics
Innovations,
6(10),
20
–
40.
https://doi.org/10.37547/tajmei/volume06issue10-
04.
I. Taj and N. Z. Jhanj
hi, “Towards Industrial Revolution
5.0 and Explainable Artificial Intelligence: Challenges
and
Opportunities.,”
International
Journal of
Computing and Digital Systems, vol. 12, no. 1, pp. 285
–
311,
Jul.
2022,
doi:
https://doi.org/10.12785/ijcds/120124.
Z. H
. Kilimci et al., “An Improved Demand Forecasting
Model Using Deep Learning Approach and Proposed
Decision Integration Strategy for Supply Chain,”
Complexity, vol. 2019, pp. 1
–
15, Mar. 2019, doi:
https://doi.org/10.1155/2019/9067367.
E. B. Tirkolaee, S. Sadeghi, F. M. Mooseloo, H. R.
Vandchali, and S. Aeini, “Application of Machine
Learning
in
Supply
Chain
Management:
A
Comprehensive Overview of the Main Areas,”
Mathematical Problems in Engineering, vol. 2021, no.
1,
pp.
1
–
14,
2021,
doi:
https://doi.org/10.1155/2021/1476043.
R. Abduljabbar, H. Dia, S. Liyanage, and S. A. Bagloee,
“Applications of Artificial Intelligence in Transport: An
Overview,” Sustainability, vol. 11, no. 1, p. 189, Jan.
2019, doi: https://doi.org/10.3390/su11010189.
Y. Issaoui, A. Khiat, K. Haricha, A. Bahnasse, and H.
Ouajji, “An Advanced System to Enhance and Optimize
Delivery Operations in a Smart Logistics Environment,”
IEEE Access, vol. 10, pp. 6175
–
6193, 2022, doi:
https://doi.org/10.1109/access.2022.3141311.
A. T. Keleko, B. Kamsu-Foguem, R. H. Ngouna, and A.
Tongne, “Artificial intelligence and real
-time predictive
maintenance in industry 4.0: a bibliometric analysis,”
AI
and
Ethics,
vol.
2,
Mar.
2022,
doi:
https://doi.org/10.1007/s43681-021-00132-6.
P. Tanuska, L. Spendla, M. Kebisek, R. Duris, and M.
Stremy, “Smart Anomaly Detection and Prediction for
Assembly Process Maintenance in Compliance with
Industry 4.0,” Sensors, vol. 21, no. 7, p. 2376, Jan. 2021,
doi: https://doi.org/10.3390/s21072376.
Ovidiu Vermesan et al., “AI and I
IoT-based Predictive
Maintenance System for Soybean Processing,” River
Publishers eBooks, pp. 327
–
352, Sep. 2022, doi:
https://doi.org/10.1201/9781003337232-27.
Mustafa Can Camur, Sandipp Krishnan Ravi, and S.
Saleh, “Enhancing supply chain resilience: A m
achine
learning approach for predicting product availability
dates under disruption,” Expert systems with
applications, vol. 247, pp. 123226
–
123226, Aug. 2024,
doi: https://doi.org/10.1016/j.eswa.2024.123226.
S. Ren, H.-
L. Chan, and T. Siqin, “Demand forec
asting in
retail operations for fashionable products: methods,
practices, and real case study,” Annals of Operations
Research, vol. 291, no. 1, Jan. 2019, doi:
https://doi.org/10.1007/s10479-019-03148-8.
Harish
Narne,
“OPTIMIZING
SUPPLY
CHAIN
MANAGEMENT
WITH
MACHINE
LEARNING
ALGORITHMS,”
INTERNATIONAL
JOURNAL
OF
ADVANCED
RESEARCH
IN
ENGINEERING
AND
TECHNOLOGY (IJARET), vol. 12, no. 03, pp. 979
–
991,
2021, Accessed: Oct. 02, 2024. [Online]. Available:
https://iaeme-
library.com/index.php/IJARET/article/view/IJARET_12_
03_091
A. N. Tarekegn, M. Ullah, and F. A. Cheikh, “Deep
Learning for Multi-Label Learning: A Comprehensive
Survey,”
arXiv.org,
Mar.
03,
2024.
https://arxiv.org/abs/2401.16549
M. A. Al-
Ebrahim, S. Bunian, and A. A. Nour, “Recent
Machine-Learning-Driven
Developments
in
E-
Commerce:
Current
Challenges
and
Future
Perspectives,” Engineered Science, vol. Volume 28 April
2024, no. 0, pp. 1044-, Dec. 2023, Available:
https://www.espublisher.com/journals/articledetails/1
044/
S. Pal et al., “Optimizing Multi
-GPU Parallelization
Strategies for Deep Learning Training,” IEEE Micro, vol.
39,
no.
5,
pp.
91
–
101,
Sep.
2019,
doi:
https://doi.org/10.1109/mm.2019.2935967.
P. Pant, A. Sai Sabitha, T. Choudhury, and P. Dhingra,
“Multi
-label Classification Trending Challenges and
Approaches,”
pp.
433–
444,
Nov.
2018,
doi:
https://doi.org/10.1007/978-981-13-2285-3_51.
K. Yin, C. Liu, A. Mostafavi, and X. Hu, “CrisisSense
-LLM:
Instruction Fine-Tuned Large Language Model for Multi-
label Social Media Text Classification in Disaster
Informatics,”
arXiv.org,
2024.
https://arxiv.org/abs/2406.15477 (accessed Oct. 02,
2024).
J. Bogatinovski, L. Todorovski, S. Džeroski, and D. Kocev,
“Comprehensive comparative study of multi
-label
classification
methods,”
Expert
Systems
with
Applications, vol. 203, p. 117215, Oct. 2022, doi:
https://doi.org/10.1016/j.eswa.2022.117215.
R. Malhotra and S. Kamal, “An empirical study to
The American Journal of Engineering and Technology
34
https://www.theamericanjournals.com/index.php/tajet
The American Journal of Engineering and Technology
investigate oversampling methods for improving
software defect prediction using imbalanced data,”
Neurocomputing, vol. 343, pp. 120
–
140, May 2019,
doi: https://doi.org/10.1016/j.neucom.2018.04.090.
J. Armando, Vicente Figueroa Fernández, A. Jiménez,
and A. Vázquez, “INDUSTRY 4.0 –
REFERENCE
FRAMEWORK AND IMPLICATIONS FOR THE CURRENT
INDUSTRY (INDUSTRIA 4.0
–
MARCO DE REFERENCIA E
IMPLICACIONES PARA LA INDUSTRIA ACTUAL),” Pistas
Educativas, vol. 40, no. 132, 2019, Accessed: Oct. 02,
2024.
[Online].
Available:
https://pistaseducativas.celaya.tecnm.mx/index.php/
pistas/article/view/1891
Ş. Scrieciu et al., “An Inquiry into Model V
alidity When
Addressing Complex Sustainability Challenges,”
Complexity, vol. 2022, p. e1193891, Sep. 2022, doi:
https://doi.org/10.1155/2022/1193891.
R. G. G. Caiado, L. F. Scavarda, L. O. Gavião, P. Ivson, D.
L. de M. Nascimento, and J. A. Garza-
Reyes, “A
fuzzy
rule-based industry 4.0 maturity model for operations
and supply chain management,” International Journal
of Production Economics, vol. 231, p. 107883, Jan.
2021, doi: https://doi.org/10.1016/j.ijpe.2020.107883.
M. Neagoe, H.-H. Hvolby
, and P. Turner, “Why are we
still queuing? Exploring landside congestion factors in
Australian bulk cargo port terminals,” Maritime
Transport Research, vol. 2, p. 100036, 2021, doi:
https://doi.org/10.1016/j.martra.2021.100036.
R. Bemthuis, W. Wang, Maria-Eugenia Iacob, and Paul
J.M. Havinga, “Business rule extraction using decision
tree machine learning techniques: A case study into
smart returnable transport items,” Procedia Computer
Science, vol. 220, pp. 446
–
455, Jan. 2023, doi:
https://doi.org/10.1016/j.procs.2023.03.057.
M. Er Kara, S. Ü. Oktay Fırat, and A. Ghadge, “A data
mining-based framework for supply chain risk
management,” Computers & Industrial Engineering,
Dec.
2018,
doi:
https://doi.org/10.1016/j.cie.2018.12.017.
T. Goswami, “Machine learni
ng behind classification
tasks in various engineering and science domains,”
Elsevier eBooks, pp. 339
–
356, Jan. 2020, doi:
https://doi.org/10.1016/b978-0-12-819443-0.00016-
7.
T.-T. Huynh-Cam, L.-
S. Chen, and H. Le, “Using Decision
Trees and Random Forest Algorithms to Predict and
Determine
Factors Contributing
to
First-Year
University
Students’
Learning
Performance,”
Algorithms, vol. 14, no. 11, p. 318, Oct. 2021, doi:
https://doi.org/10.3390/a14110318
Z. Sun, G. Wang, P. Li, H. Wang, M. Zhang, and X. Liang,
“An improved random forest based on the
classification accuracy and correlation measurement of
decision trees,” Expert Systems with Applications, vol.
237,
p.
121549,
Mar.
2024,
doi:
https://doi.org/10.1016/j.eswa.2023.121549.
A. Khan, A. Sohail, U. Zahoo
ra, and A. S. Qureshi, “A
survey of the recent architectures of deep convolutional
neural networks,” Artificial Intelligence Review, vol. 53,
Apr. 2020, doi: https://doi.org/10.1007/s10462-020-
09825-6.
M. Umer et al., “Impact of convolutional neural networ
k
and FastText embedding on text classification,”
Multimedia Tools and Applications, Aug. 2022, doi:
https://doi.org/10.1007/s11042-022-13459-x.
Jahin, Md Abrar, A. Shahriar, and M. A. Amin, “MCDFN:
Supply Chain Demand Forecasting via an Explainable
Multi-Channel Data Fusion Network Model Integrating
CNN,
LSTM,
and
GRU,”
arXiv.org,
2024.
https://arxiv.org/abs/2405.15598 (accessed Oct. 02,
2024).
Z. Zhang, F. Feng, and T. Huang, “FNNS: An Effective
Feedforward Neural Network Scheme with Random
Weights for Processing Large-
Scale Datasets,” Applied
Sciences, vol. 12, no. 23, p. 12478, Dec. 2022, doi:
https://doi.org/10.3390/app122312478.
A. Manno, M. Intini, O. Jabali, F. Malucelli, and D. Rando,
“An ensemble of artificial neural network models to
forecast hour
ly energy demand,” Optimization and
Engineering,
Mar.
2024,
doi:
https://doi.org/10.1007/s11081-024-09883-7.
J. Shukla, M. Barreda-Angeles, J. Oliver, G. C. Nandi, and
D. Puig, “Feature Extraction and Selection for Emotion
Recognition from Electrodermal Ac
tivity,” IEEE
Transactions on Affective Computing, pp. 1
–
1, 2019,
doi: https://doi.org/10.1109/taffc.2019.2901673.
