International Journal of

INTELLIGENT SYSTEMS AND APPLICATIONS IN

ENGINEERING
ISSN:2147-67992147-6799 www.ijisae.org Original Research Paper

Automating Daily Task in Manufacturing and Production Sites Via

Machine Learning Intelligence
1
K. Swarupa Rani, 2M. Kumarasamy, 3Balachandra Pattanaik, 4Sharath Kumar Jagannathan, 5Manjula
Pattnaik, 6B. Girimurugan
Submitted: 02/10/2023 Revised: 21/11/2023 Accepted: 01/12/2023

Abstract: Robots have been utilized for some time now in the manufacturing industry. These robots work safely alongside humans and
gain knowledge from their interactions with them. However, there are still sectors of the economy that are hesitant to adopt robots for a
variety of reasons, including those pertaining to technology and the economy. The advancement of robotics has led to the creation of
capabilities that span a greater breadth of applications than those that were previously utilized. In this paper, we develop an auto encoder
based modelling to automate the daily task using robotic process automation in manufacturing and production sites. The unsupervised
learning models achieves better processing of the automation and provides better accurate results than the existing methods. The software
is able to perform simultaneous analysis on a number of logs, which enables it to discover processes and variations that were previously
unknown. The platform has the potential to legitimately outperform technology-based businesses when it comes to the automation of
processes, whether those processes are carried out in a physical or digital environment.

Keywords: Task, Manufacturing, Production Sites, Machine Learning, Artificial Intelligence.

1. Introduction they have learnt, they infer new knowledge about the
world. When applied to ML models, the issues that surface
It is essential to adopt strong software engineering (SE)
as a direct result of employing erroneous information
principles and techniques in order to construct machine
stand in stark contrast to the kinds of issues that are seen
learning (ML) software systems that are relevant in the
in conventional programming. Not only does this result in
real world. This is one of the goals of this project. Methods
erroneous conclusions for that specific set of data, but it
and approaches, which are supported by a wide variety of
can also be used to support the building of a model that is
tools, are at the center of both the theoretical
incorrect or insufficient in general. Additionally, even if
underpinnings and the practical applications of software
the model generates findings that are adequately accurate,
engineering (SE). These were designed in order to
its performance will substantially decline over the course
guarantee the methodical design of dependable systems.
of time regardless of how well it was originally designed.
During the development process of ML systems, other
components that are not a part of ML models are taken Throughout the process of designing and maintaining
into consideration [1]. software systems that are powered by machine learning [2,
3, 4], it is normal for there to be data inaccuracies, and it
In order for machine learning models to be trained, high-
can be difficult to identify these flaws. Errors in data can
quality input data is required. After that, these models look
result in considerable financial losses for organizations.
for relevant patterns in the data, and based on the patterns
To give you an example, LinkedIn experienced losses and
1
Assistant Professor, Department of IT, PVP Siddhartha Institute of
was required to invest a large amount of work in order to
Technology, Kanuru, Vijayawada, Andhra Pradesh, India.
2
Assistant Professor, Department of Computer Science, College of expose data flaws in their job suggestion platform [5].
Engineering and Technology, Wallaga University, Nekemte, Ethiopia, Error data can be caused by a number of factors, including
Africa.. a lack of knowledge of complicated data relationships,
3
Professor, Department of Electrical and computer engineering, College
defects in application code, data drift from sensors, and
of Engineering and Technology, Wallaga University, Nekemte, Ethiopia,
Africa. data gaps induced by disruptions in network connections
4
Assistant Professor, Frank J. Guarini School of Business, Saint Peter's [6, 7, 8].
University, Jersey City, NJ-07306, USA.
5
Associate Professor, College of Business Administration, Princess When doing an analysis to determine the quality of the
Nourah Bint Abdulrahman University, Riyadh, KSA. data, it is imperative to take into consideration the data
6
Assistant Professor, KL Business School, Koneru Lakshmaiah precision, completeness, consistency, and timeliness [9 -
Education foundation (Deemed to be University), Andhra Pradesh, India.
1 2
13]. Data validation procedures really come into their own
swarupapvpsit@gmail.com, drmksamy115@gmail.com,
3
balapk1971@gmail.com, 4
sjagannathan@saintpeters.edu, when working with extremely large amounts of data on a
5
drmanjula23@gmail.com, 6skbgiri@yahoo.co.in consistent basis. In order to assess the performance of
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 47
machine learning (ML) software, a procedure that is the collaborative assembly process, the modules
conceptually analogous to this one is called the data responsible for power, protection, and interface have all
validation method [14]. been described. During this process, humans and robots
will collaborate in close proximity to one another in an
Even though there are a number of problems that can be
open setting.
discovered in the currently available data validation tools,
such as flaws in the implementation and uncoupling from 2. Related Works
the data cleaning capabilities, a significant amount of
With the support of the strategic framework that
emphasis is placed on the implementations of the various
automation provides, many automation technologies can
tools.
be adopted independently of one another. In order to
Despite the fact that these tools are currently under accomplish this goal, it is necessary to locate tasks,
development, this remains the case. Documentation possess the flexibility necessary to reuse automated
regarding adoption experiences with the data validation operations, and make the most of the capabilities offered
procedure is produced exceptionally infrequently. The by the automated system [18]. The goal of automating a
members of your team who are just starting to put ML- process is to make it more effective while also reducing
enabled software solutions into production could stand to costs. This can be accomplished through the
benefit greatly from the lessons that are taught in this implementation of a number of different strategies,
article. Engineering demands a significant time and including the standardization of routine tasks and the
financial investment in order to develop and maintain a enhancement of the utilization of data received digitally.
validation [15]. This can be accomplished by reducing the amount of time
Developing and maintaining a data validation method and that is wasted and making more efficient use of the data
instrument since established protocols does not exist for that has been acquired [19].
building a validation process, there is no way to guarantee These data can assist organizations in making decisions
that the overall quality of the data will be high. The vast that are both more informed and more timely. Businesses
majority of engineering teams choose not to incorporate can, among other things, grow, become more integrated,
data validation tools into their processes if those tools do and operate more efficiently as a result of automation
not fulfill particular requirements [16]. The primary capabilities [20]. Our comprehension of the potentials and
reason for this is the difficulties that is involved in confines offered by RPA technologies has been expanded,
carrying out the task. This would be a concern, for and clear answers to the issues raised by these
instance, if it caused workflows to produce obsolete data technologies have been supplied. separation from
as a result of higher wait times [17]. frameworks whose sole purpose is to concentrate on the
It is now appropriate for use in modern services in production of automation tools or concepts. The following
addition to taking on responsibilities that are significantly is a review of the many key capabilities of automation that
more complicated. Peripheral components of a system are can be used to improve automation in industries. This
more amenable to the incorporation of new technology topic is shown in Table 1.
than central components. During the current iteration of
Table 1. Capabilities of Task Automation in Buildings
S. No Capabilities
1 Automate business
2 Increase human skill
3 Scalability of operations
4 Gain flexibility
5 Improve operational efficiency
6 Detect risk
7 Efficiently functioning
8 Automate repeat operation
9 Speed the digital process

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 48
 10 Automated Billing
11 Automated workload

If organizations are going to be able to fully automate their to accomplish this goal, it makes use of a vast array of
processes, the technology of today has to be able to machine learning, pre-packaged software, and labor
overcome the limitations imposed by older approaches. automation technologies [22].
Companies now have the ability to expand beyond the
The only way for the lives of workers to get better is for
limitations of certain processes and find the most effective
their company to become more successful and have higher
methods for automating a wide variety of high-level, high-
levels of production. Automation in the manufacturing
volume jobs. This is made possible by advances in
business is made possible by employing a wide range of
information technology. Automation provides businesses
digital technology in a variety of different applications.
with advanced analytical tools to assist such businesses in
Even if automation has been utilized extensively for a
overcoming the limitations of relying solely on a single
considerable amount of time, it is currently of the utmost
data collecting platform [21].
significance to adapt to the nature of automated systems,
Instead of adding artificial intelligence, it enables business which is constantly evolving. Despite the fact that
customers to automate the entire process of jobs at once. automation has been widely utilized for a considerable
This is a significant time saver. For the automation amount of time, this remains the case [23].
process to be successful, it is essential that the existing
3. Proposed Method
information technology infrastructure and business
processes of the company continue to function normally. The application of machine learning can result in
The gap that previously existed between traditional improvements that are optimized not just in terms of
computer systems and automation has been bridged by a efficiency but also in terms of quality as illustrated in
technology known as robotic process automation. In order figure 1.

Robotic Robotic
Machine
Desktop Process AI
Learning
automation Automation

Fig 1: Process of Automation

Manual Digital Perspective Decision Deductive

Intervention Triggers Analytics Engine Analytics

Fig 2: Role of Manual and user intervention in Automation

Analysis using
Model

Increasing Product
Quality and Target Adaptation
Productivity

Fig 3: Proposed Modelling

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 49
 Business Discovery

Select Process

Create Rules

Evaluation

Deployment

Monitoring Changes

Fig 4a: Process of Automation

The ability of RPA, to process unstructured data inputs by implementation process is illustrated in figure 2, which
capitalizing on powerful AI capabilities considerably can be found here. As can be seen in Figure 3a and b of
increases the automation and broadens the scope of its the implementation pipeline for cognitive robotic process
potential applications. The RPA lifecycle provides a automation, the cognitive capabilities of the system to
structure for the automation process, acting as a learn and make decisions. This is an important step in the
framework that enables quality control at each stage of the development of cognitive RPA.
lifecycle. The usual manner of carrying out the

Business Discovery

Data Analysis

Data Preparation

Train AE Model

Evaluation

Deployment

Obtain Feedback

Fig 4b: Proposed Process of Automation with AE

Auto-encoders, also known as AEs, are a well-known representations. AEs are abbreviated as AE. Auto-
unsupervised learning method that may be applied to the encoders are typically applied when working with data
process of instructing neural networks how to learn that has a high number of dimensions, and dimensionality
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 50
reduction provides an explanation for how the data in attributes from data that has not been labeled, and its name
question is represented. Both of these processes are comes from the fact that this method of learning is
referred to as dimensionality reduction. unsupervised.
Autoencoders are constructed with an encoder, a code, Figure 4 depicts the architecture of an autoencoder. As can
and a decoder all incorporated into its design. The input is be seen in the figure, the architecture of an autoencoder
compressed, and the encoder produces a code; the consists of both an encoder and a decoder. After being
component that is responsible for decoding the code is processed by the encoder, the data are then output as low-
called the decoder. In recent years, generative data models dimensional latent vectors. The encoding process begins
have been taught with the assistance of artificial with the data being fed into the device. The data from the
engineers. output stream are sent via the decoder, which makes use
of those latent vectors in order to recreate the data from
Autoencoders are networks that strive to encode input into
the input stream. Data compression, denoising, and
a latent space and decode it subsequently. This is how they
anomaly detection are just a few examples of significant
got their name. Unsupervised learning is a form of
applications.
learning that permits the extraction of generally beneficial

Reconstructed
Input Data Encoder Latent Vectors Decoder
Output

Fig 5: Auto Encoder Process

The operation of an autoencoder starts with the function u(x,y) to the input anxnin, is applied using max
compression of the input data into latent vectors (which pooling that minimizes the size of the input by substituting
are not displayed), as shown in Figure 5, and is then each neighborhood with the maximum value aout. This
followed by the re-decoding of the vectors into the starting results in a smaller amount of input data. This results in a
data. This may be seen in action in the figure. reduction in the size of the input.
An autoencoder that is built up of convolutional layers, For a variety of different causes, one begins to suffer
also known as a Conv autoencoder, is utilized in place of degradation problems when deeper networks begin to
a traditional autoencoder such as a fully connected converge. This difficulty becomes more severe as the
autoencoder since the Conv autoencoder is superior at network depth increases, which ultimately leads to a drop
managing the data associated with 2D images. In case you in accuracy. It is difficult for autoencoders to gain an
were wondering what a convolutional autoencoder looks understanding of the circumstance based on the data when
like, this is an example of one. Evolvable filters are used they are presented with this obstacle. As can be seen in
to construct the layers of parameters that make up Figure 5, in order to remedy this problem, we put in place
convolutional neural networks (CNNs), which are skip links between the encoder and decoder levels.
becoming increasingly popular for their application in the Because of the skip connections that are present between
interpretation of visual input. the encoders and decoders of each layer, it is possible for
the issue of pixel-wise prediction to converge to a solution
CNNs couldn't function without the indispensable conv.
that is more optimal. X1 will be used to represent the
An activation function, such as the rectified linear unit
outputs of the encoder layer, and X2 will be used to
(ReLU) or the sigmoid function, is responsible for the
represent the outputs of the decoder layer. As the input to
generation of a two-dimensional feature map after the
the succeeding decoder layer, you should make use of the
input data has been processed by the Conv. In addition,
formula that is listed below:
the resistance that exists between the filter and the data
that is being input is computed by making use of the data F(X1,X2)=X1⊕X2 (2)
that is being input.
The network makes use of skip connections so that the
aout=max(an×ninu(n,n)) (1) various components of the feature maps that are utilized
by the encoder and the decoder can be combined together.
The spatial continuity of the feature map must be
These maps are used by both the encoder and the decoder.
preserved in order for the pooling layer to fulfill its
This facilitates the process of recovering the picture in
primary function, which is to concurrently reduce the
some way. During the course of this research, an
resolution of the map. This results in learning that is both
autoencoder was utilized and instructed to learn how to
more quick and more compact, while also minimizing the
amount of processing overhead. When the window
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 51
reconstruct normal images from ones that had been Using Eq. (2) as a guide, decode a function gθ that
tampered with. transforms the input space into the feature space. This
function transforms the input space into the feature space.
The table encoder and decoder portions make up the upper
and lower tiers, which are situated above and below the r=gθ(h) (4)
table central horizontal line, respectively. The part of the
Probabilistic models are constructed and trained on the
system known as the encoder is responsible for converting
basis of a probability function that has been provided in
encoder data, which has a size of 400x400x3, into latent
order to maximize (roughly) data similarity. The data are
vectors. In the meantime, the component known as the
compared to the function in order to accomplish this. The
decoder is responsible for producing decoder data, which
process of teaching autoencoders to decode data can take
is also of size 400x400x3 in its entirety. This encoder
place in a variety of different ways. Both the encoder and
makes it feasible for the defect image to be reconstructed
the decoder have had their respective parameter sets
as a defect-free image during the decoding process so that
trained so that they can produce the same primary input.
it can be displayed correctly. After that, the image that was
This guarantees that the product they create will be the
generated can be utilized during the process of defect
same. In the process of reconstruction, every effort is
detection image subtraction.
made to leave as little room for error as is humanly
Both batch normalizing and the ReLU activation function possible. Applying the equation results in the
are utilized in each and every convolutional layer that is determination of the reconstruction error, which is
present. The first step is to do batch normalization. Models denoted by the symbol L(x,r). It provides a numerical
are able to rapidly assimilate new information thanks to value for the difference between the original version of x
the ReLU activation function, which enables them to learn and the reconstructed version of x, which is denoted by r.
properties from data that are significant both
L(x,r)=‖x−r‖2 (5)
quantitatively and qualitatively. Each skip connection
combines the conversion output from the encoder with the Training an autoencoder can be summed up in a few words
upsampling output in order to make up for any data that as the process of locating the parameter vector θ that
may have been lost as a result of the compression process results in the least amount of reconstruction error is
that took place in the encoder. This is done in order to applied, where x(t) represents a training example. This can
compensate for any data that may have been lost as a result be said to be the process of minimizing the amount of error
of the compression process. that occurs during reconstruction.

It is possible to disassemble the autoencoding process into JAE(θ)=∑L(x(t),gθ(fθ(x(t)))) (6)

its component parts, which are as follows: To accomplish the goal of lowering the value in question,
(a) Encoder the stochastic gradient descent method is utilized the vast
majority of the time. This method is conceptually
Encoder, a function for the extraction of features that
comparable to the multilayer perceptron training
generates a feature vector based on the inputs that are
approach.
provided; this function is part of the feature extraction
process. Since this is the case, we may define the encoding 4. Results and Discussions
datasets as x(t) and the encoder function as fθ, which will
For the experiments, the following of hardware are
give us the following equation (3):
utilized, which is shown in Table 1. Before the NN can be
h(t)=fθ(x(t)),x(t)={x(1),…,x(T)} (3) trained, there must first exist a dataset that it can draw
information from. To the best of our knowledge, there is
where h(t) - features vector.
no dataset that is open to the public that is concentrated
(b) Decoder solely on user interfaces. We were able to confirm this by
looking into the matter.
Table 1: Hardware Specification
S.No Specification
1 4 Terabyte Storage in build Capacity
2 Intel i5
3 NVIDIA Graphics Card
4 8 Terabyte Memory

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 52
Therefore, the researchers who were responsible for this dataset can be broken down into three distinct
study are the ones who should be held accountable for the subcategories: training, validation, and testing. Each of
creation of the dataset that was used in their work. The these subcategories can be extracted independently.

450
400
350
300
Time (ms) 250
200
150
100
50
0
ANN CNN AE Proposed AE
Algorithms

Fig 6: Training time

35

30

25
Time (ms)

20

15

10

0
ANN CNN AE Proposed AE
Algorithms

Fig 7: Testing time

1.6
1.4
1.2
1
Error %

0.8
0.6
0.4
0.2
0
ANN CNN AE Proposed AE
Algorithms

Fig 8: Training – MAE

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 53
 3

2.5

Error %
1.5

0.5

0
ANN CNN AE Proposed
AE
Algorithms

Fig 9: Testing - MAE

2.5

1.5
Error %

0.5

0
ANN CNN AE Proposed
AE
Algorithms

Fig 10: Training – RMSE

2.5

2
Error %

1.5

0.5

0
ANN CNN AE Proposed AE
Algorithms

Fig 11: Testing - RMSE

If companies want to keep up with their current level of increase in overall quality. In the event that certain quality
success in the market, their products will require an standards are not met, either the production line will be
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 54
brought to a end, or the finished product will fall short of 5. Conclusions
the requirements set forth by the customer. The concept of
The purpose of automation is to enhance sustainability
consistent takes on a different connotation depending on
while simultaneously increasing savings and generating
the kind of product that is being discussed.
additional income. The achievement of this objective is
This concept of consistency can be extended to cover the made possible through the integration of automated
entirety of the manufacturing process by including technology and instruments with human labor. Businesses
production factors such as the quantity of raw materials have the opportunity to educate themselves on the various
used, the amount of time it takes to manufacture the methods of automation, how these methods relate to one
component, and the number of people who are needed to another, and how these methods can be aggregated and
finish it. This would bring the concept of consistency in facilitated in order to acquire the appropriate tools and put
line with industry standards. When one is attempting to them to use in their operations. It is necessary for software
maximize a certain set of output factors, it can be difficult to be able to communicate with one another as the use of
to learn everything about a manufacturing process that is automated processes becomes more widespread. In
necessary to do so in order to achieve that goal. Existing addition, the development of new tools that enable plug-
systems are already optimized for the precise execution of and-play architectures could be of assistance to businesses
establishing criteria; however, self-optimizing systems in their efforts to successfully scale up their operations.
should be used in order to achieve the highest possible The term automation refers to the process of fully
level of output consistency. automating a process by putting in place a toolchain.
It is necessary to have professional knowledge of the When using this approach, companies are able to automate
mechanisms and boundary conditions that execute each not just a portion of their business processes but rather all
phase of the manufacturing chain in order to have a of those processes altogether.
complete understanding of this concept. Only then will References
you be able to comprehend it in its entirety. They consider
it to be standard practice to modify already-existing [1] Baduge, S. K., Thilakarathna, S., Perera, J. S.,
content in order to incorporate newly developed features, Arashpour, M., Sharafi, P., Teodosio, B., ... &
and they see it as perfectly acceptable to do so. Every new Mendis, P. (2022). Artificial intelligence and smart
generation of manufacturing machinery and control vision for building and construction 4.0: Machine
technologies brings with it improvements in both the and deep learning methods and
effectiveness and dependability of production systems. applications. Automation in Construction, 141,
This trend is expected to continue. The manufacturers use 104440.
materials that are more resilient and incorporate quicker [2] Dash, R., McMurtrey, M., Rebman, C., & Kar, U. K.
actuators wherever it is feasible to do so. It is possible to (2019). Application of artificial intelligence in
achieve better control of the circuit and permit faster automation of supply chain management. Journal of
coordination with the unit through the implementation of Strategic Innovation and Sustainability, 14(3), 43-
faster bus systems that are equipped with sensors. This 53.
makes it possible to achieve both of these goals.
[3] Howard, J. (2019). Artificial intelligence:
In particular, the results of our tests indicate that statistical Implications for the future of work. American
learning in latent spaces leads to fewer errors on average Journal of Industrial Medicine, 62(11), 917-926.
and can achieve satisfactory performance with a reduced
[4] Rahimian, F. P., Seyedzadeh, S., Oliver, S.,
number of data points. Additionally, our findings suggest
Rodriguez, S., & Dawood, N. (2020). On-demand
that this type of learning can be accomplished with a
monitoring of construction projects through a game-
smaller number of data points. It is a well-known fact that
like hybrid application of BIM and machine
the process of reinforcement learning in latent spaces is
learning. Automation in Construction, 110, 103012.
both noticeably quicker and more reliable than in other
settings. In each and every experiment that we carried out, [5] Iqbal, R., Maniak, T., Doctor, F., & Karyotis, C.
we found that learning in the AE-based latent space (2019). Fault detection and isolation in industrial
performed noticeably better than the methods that are processes using deep learning approaches. IEEE
currently considered to be state-of-the-art. In comparison Transactions on Industrial Informatics, 15(5), 3077-
to DAE, the results obtained are consistent with the idea 3084.
that deep autoencoder neural networks have superior
[6] Woschank, M., Rauch, E., & Zsifkovits, H. (2020).
performance in terms of approximation.
A review of further directions for artificial

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 55
 intelligence, machine learning, and deep learning in [16] McClellan, M., Cervelló-Pastor, C., & Sallent, S.
smart logistics. Sustainability, 12(9), 3760. (2020). Deep learning at the mobile edge:
Opportunities for 5G networks. Applied
[7] Madakam, S., Holmukhe, R. M., & Jaiswal, D. K.
Sciences, 10(14), 4735.
(2019). The future digital work force: robotic
process automation (RPA). JISTEM-Journal of [17] Angelopoulos, A., Michailidis, E. T., Nomikos, N.,
Information Systems and Technology Trakadas, P., Hatziefremidis, A., Voliotis, S., &
Management, 16. Zahariadis, T. (2019). Tackling faults in the industry
4.0 era—a survey of machine-learning solutions and
[8] Waring, J., Lindvall, C., & Umeton, R. (2020).
key aspects. Sensors, 20(1), 109.
Automated machine learning: Review of the state-
of-the-art and opportunities for healthcare. Artificial [18] González, G., & Evans, C. L. (2019). Biomedical
intelligence in medicine, 104, 101822. Image Processing with Containers and Deep
Learning: An Automated Analysis Pipeline: Data
[9] Sircar, A., Yadav, K., Rayavarapu, K., Bist, N., &
architecture, artificial intelligence, automated
Oza, H. (2021). Application of machine learning and
processing, containerization, and clusters
artificial intelligence in oil and gas
orchestration ease the transition from data
industry. Petroleum Research.
acquisition to insights in medium‐to‐large
[10] Wu, J., Cai, N., Chen, W., Wang, H., & Wang, G. datasets. BioEssays, 41(6), 1900004.
(2019). Automatic detection of hardhats worn by
[19] Kumar, Y., Kaur, K., & Singh, G. (2020, January).
construction personnel: A deep learning approach
Machine learning aspects and its applications
and benchmark dataset. Automation in
towards different research areas. In 2020
Construction, 106, 102894.
International conference on computation,
[11] Park, S. T., Li, G., & Hong, J. C. (2020). A study on automation and knowledge management
smart factory-based ambient intelligence context- (ICCAKM) (pp. 150-156). IEEE.
aware intrusion detection system using machine
[20] Yuvaraj, N., Karthikeyan, T., & Praghash, K. (2021).
learning. Journal of Ambient Intelligence and
An improved task allocation scheme in serverless
Humanized Computing, 11(4), 1405-1412.
computing using gray wolf Optimization (GWO)
[12] Antonopoulos, I., Robu, V., Couraud, B., Kirli, D., based reinforcement learning (RIL) approach.
Norbu, S., Kiprakis, A., ... & Wattam, S. (2020). Wireless Personal Communications, 117(3), 2403-
Artificial intelligence and machine learning 2421.
approaches to energy demand-side response: A
[21] Popescu, G. H., Petreanu, S., Alexandru, B., &
systematic review. Renewable and Sustainable
Corpodean, H. (2021). Internet of things-based real-
Energy Reviews, 130, 109899.
time production logistics, cyber-physical process
[13] Varma, M., Lu, M., Gardner, R., Dunnmon, J., monitoring systems, and industrial artificial
Khandwala, N., Rajpurkar, P., ... & Patel, B. N. intelligence in sustainable smart
(2019). Automated abnormality detection in lower manufacturing. Journal of Self-Governance &
extremity radiographs using deep learning. Nature Management Economics, 9(2).
Machine Intelligence, 1(12), 578-583.
[22] Yuvaraj, N., Srihari, K., Dhiman, G.,
[14] Ashima, R., Haleem, A., Bahl, S., Javaid, M., Mahla, Somasundaram, K., Sharma, A., Rajeskannan, S. M.
S. K., & Singh, S. (2021). Automation and G. S. M. A., ... & Masud, M. (2021). Nature-
manufacturing of smart materials in Additive inspired-based approach for automated
Manufacturing technologies using Internet of Things cyberbullying classification on multimedia social
towards the adoption of Industry 4.0. Materials networking. Mathematical Problems in Engineering,
Today: Proceedings, 45, 5081-5088. 2021.
[15] Peters, E., Kliestik, T., Musa, H., & Durana, P. [23] Amershi, S., Begel, A., Bird, C., DeLine, R., Gall,
(2020). Product Decision-Making Information H., Kamar, E., ... & Zimmermann, T. (2019, May).
Systems, Real-Time Big Data Analytics, and Deep Software engineering for machine learning: A case
Learning-enabled Smart Process Planning in study. In 2019 IEEE/ACM 41st International
Sustainable Industry 4.0. Journal of Self- Conference on Software Engineering: Software
Governance & Management Economics, 8(3). Engineering in Practice (ICSE-SEIP) (pp. 291-300).
IEEE

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(7s), 47–56 | 56