DL- based model with optimization algorithm for pandemic detection.

Ref Study goal DL technique Datase Optimis Results Deductions others

used t ation
algorith
m
Predictive ARIMA, SARIMA, Enhanced CNN-LSTM- The framework focused on forecasting temperature time
(Selm Analytics
Framework.
LSTM, CNN-
LSTM.
analysis
and
based DL
model
series data using traditional and DL predictive analytics
methods. The problem now is how to benefit from all of
y, prediction exhibits this data gathered by sensing and monitoring changes
Moha accuracy higher like temperature, light, and position.
closeness to
med, actual Autoregressive Integrated Moving Average (ARIMA),
& values
compared to
Medh LSTM
at,
2024)
(Li, Mechanical
Properties of
CNN Resulting
stress-
More useful
and
However, recycled composites cannot be widely used
owing to their poor reliability, which is caused by the
Wang, Recycled strain dependable random and significant variations in their mechanical
Zhao, Composites curves recycled properties. Surface images of the resulting specimens
were composite were captured and tension tests subsequently conducted
& examined materials to obtain their mechanical properties.
Chen,
2024)
Shekar Rainfall-Runoff ANN, KNN, CNN-RNN The study The findings suggest that the CNN-RNN model could be a
P.R.; Modeling XGBoost, RF, model concluded valuable tool for various applications related to
Mathew CNN, CNN-RNN showed that the sustainable water resource management, flood control,
A.; superior CNN-RNN and environmental planning.
Yeswanth performan model The years 2003–2007 are classified as the calibration or
P.V.; ce in outperforme training period, while the years 2008–2009 are classified
Deivalak simulating d other as the validation or testing period for the span of time
shmi S monthly models in 2003 to 2009.
(P.R., rainfall- simulating
2024). runoff monthly
rainfall-
runoff,
based on
the collected
data
Issahaku Covid-19 Modified VGG16, High Our method We analyzed the performance of the proposed
F.-L.Y.; Detection Faster R-CNN efficacy in excelled in methodology using a dataset of 2059 Covid-19 positive
Liu X.; Lu detecting Covid-19 chest X-ray images and a diverse collection of 25,000
K.; Fang Covid-19 detection cough recordings, including 716 positive, 7693 negative,
X.; using using chest and 1591 symptomatic cases. The datasets were divided
Danwana chest X- X-ray into 80% for training and 20% for testing. We compared
S.B.; ray images and the results of the proposed model to already existing
Asimeng images cough networks using chest X-rays and coughs as datasets, and
E. and cough recordings, the proposed model outperformed them. Experiments
recordings surpassing also show that the trained model resists training data
. existing quantity and quality and can effectively detect Covid-19
networks in chest X-rays and coughs.
and
handling
training data
variations
effectively.
(D., AI in
Manufacturing
CNN, LSTM,
hybrid models
Enhanced
predictive
AI and DL
techniques
Through meticulous research, real-world case studies,
and expert insights, readers will embark on a
2024); maintenan enhance transformative journey, unraveling the intricacies of
Sivaraja ce, manufacturi predictive maintenance, demand forecasting, quality
M. demand ng assurance, and beyond.
forecastin processes by
g, and leveraging
quality real-time
assurance data and
in advanced
manufactu models,
ring. optimizing
efficiency,
competitive
ness, and
sustainability
.
(A.A., Intrusion
Detection
CGAN, VGG19,
AO
High
accuracy
The study
shows that
The novelty of this study is implanted in the integrated
approach, distinguishing it from previous research and
2024); System for Smart in combining contributing to effective protection against cyber threats
Ergül Ö.; Grids detecting Game in smart grids. Addressing the critical need for robust SG
Rahebi J. cyber- Theory, intrusion detection,
attacks swarm
with intelligence,
99.82% and DL
accuracy. techniques
provides a
robust
solution for
SG intrusion
detection.
(Islam Sentiment
Analysis
Capsule-based
RNN, CNN, CRDC
High
accuracy
The
innovative
Additionally, we introduced the innovative CRDC
(Capsule with Deep CNN and Bi structured RNN) model,
, et model in use of which demonstrated superior performance compared to
al., sentiment capsule- other methods. Our proposed approach achieved
analysis based RNNs remarkable accuracy across different databases: IMDB
2024). with the and the (88.15%), Toxic (98.28%), CrowdFlower (92.34%), and ER
CRDC CRDC model (95.48%). Hence, this method holds promise for
model significantly automated sentiment analysis and potential deployment.
 outperfor enhances
ming sentiment
others. analysis
accuracy.
(A. I. , Anomaly
Detection and
Various DL
models
Effective
anomaly
2024); Time Series detection
Amin R. Forecasting and time
series
forecastin
g in
banking,
healthcare
, and
manufactu
ring.
(Sobc Synthetic data,
an
It accentuates
practical
This
controlled
zak- algorithmically aspects, generatio
Micha birthed prioritizing the n ensures
counterpart to basic the
lowsk operational data, applicability of preservati
a, is the linchpin for synthetic high- on of
overcoming dimensional privacy
Borah constraints data. Each and
associated with chapter unveils a facilitates
, sensitive or facet of synthetic learning
Polko regulated data's prowess, across
information. In from its impact datasets,
wski, high-dimensional on society to its which is
& data, where the role in machine crucial
dimensions of learning when
Mishr features and applications. It dealing
a, variables often provides a with
surpass the roadmap for incomplet
2024) number of navigating the e, scarce,
available nuanced terrain or biased
observations, the of data privacy, data. Ideal
emergence of continuity, and for
synthetic data generalization researcher
heralds a through the lens s,
transformation. of synthetic data. professors
Applications of ,
Synthetic High practition
Dimensional ers,
Data delves into faculty
the algorithms members,
and applications students,
underpinning the and online
creation of readers,
 synthetic data, this book
which surpass transcend
the capabilities s
of authentic theoretical
datasets in many discourse
cases. Beyond
mere mimicry,
synthetic data
takes center
stage in
prioritizing the
mathematical
domain,
becoming the
crucible for
training robust
machine learning
models.
(A. Parking
Occupancy
ANFIS-RNN,
ANFIS-LSTM,
Superior
performan
By
integrating
ANFIS-RNN improved MSE by 30.61%, RMSE by 16.70%,
MAE by 21.21%, MAPE by 21.58%, and R2 by 1.03% over
E. , Forecasting ANFIS-GRU ce in ANFIS with RNN. ANFIS-LSTM surpassed LSTM with 34.04% MSE,
2024); parking DL 18.76% RMSE, 26.16% MAE, 27.71% MAPE
occupancy architecture improvements, and 1.04% R2 increment. ANFIS-GRU
Křupka J.;
forecastin s, the study exceeded GRU with 27.54% MSE, 14.85% RMSE, 19.27%
Jovčić S.;
g achieves MAE, 20.01% MAPE improvements, and 1.03% R2 boost.
Simic V.
compared significant
to improvemen
standalon ts in parking
e DL occupancy
models. forecasting.
(H., This work aims
to explore the
The combination
of deep
2024); fusion of DL and convolutional
Zhou M. blockchain neural networks
technology for with SVD and
research blockchain
applications in technology is
photography and currently
art studies. A exhibiting strong
digital image convergence
watermarking performance and
model is the ability to
established, extract critical
followed by the image features
development of during photo
a versatile post-processing.
watermarking It currently
algorithm model demonstrates
using singular excellent
value convergence and
decomposition holds significant
(SVD) and DL potential for
techniques, applications such
thereby as image
extending the copyright
applicability of protection and
watermarking enhancing image
technology. processing in
Furthermore, the various domains.
performance
parameters of
the blockchain
model are
designed using
the practical
byzantine fault
tolerance
consensus
mechanism to
ensure
transaction
consistency and
reliability. Finally,
the algorithm’s
normalized
cross-correlation
is analyzed and
attack
experiments are
conducted on
the original
images.
Adjustments are
made to the
model to
simulate the
scalability of
sharded
blockchains, and
scalability of
cumulative
revenue is
evaluated
through
simulation
experiments.
Through
simulation
experiments, it is
demonstrated
 that this model
can capture
value functions
in a sharded
blockchain
environment,
accelerating
convergence
speed and
achieving
scalability
(J. S. , In recent years,
metaheuristic
These results
indicate that
The
objectiv
2024); optimizers have incorporating e of this
Sandhu grown in hyperparameter paper is
J.K.; popularity due to metaheuristic to
Kumar Y. their ability to techniques is design a
efficiently able to improve system
optimize the detection as that
complex, high- well as diagnosis combin
dimensional performance of es
problems that models for skin metahe
are difficult to disease. uristic
solve using However, optimiz
conventional additional ers with
optimization research is various
techniques. In required to AI
the context of investigate new based
multiple learning strategies and classifie
models, where enhance the rs to
large-scale scalability of the detect
datasets and current methods. and
complex models diagnos
necessitate e skin
efficient disease
optimization s.
techniques,
these
optimization
algorithms are
particularly
useful.
(A.C., Human Action
Recognition
LSTM Results
demonstra
The study
validates the
However, the processing aspect is evolving towards
distributed approaches, leveraging edge-computing.
2024); (HAR) te SOTA proposed These distributed systems are capable of effectively
Losada- accuracy approach for addressing the detection of people or events at each
Gutiérrez people individual node. Most of these systems, rely on the use of
C.; detection DL and segmentation algorithms which enable them to
Marrón- and action achieve high performance, but usually with a significant
Romera recognition computational cost, hindering real-time execution. This
M.; in real-world paper presents an approach for people detection and
Gardel- video action recognition in the wild, optimized for running on
Vicente surveillance the edge, and that is able to work in real-time, in an
A.; applications. embedded platform.
Bravo-
Muñoz I.
(A. Nevertheless,
developments in
The
results
H. , digital suggest
2024); technology have that the
been used to CNN
Jafari
produce software methodolo
Navimipo
that poses a gy is the
ur N.;
threat to most often
Dag H.;
democracy, employed
Unal M.
national security, DL
and method in
confidentiality. publicatio
Deepfake is one ns.
of those DL-
powered apps
that has lately
surfaced. So,
deepfake
systems can
create fake
images primarily
by replacement
of scenes or
images, movies,
and sounds that
humans cannot
tell apart from
real ones.
Various
technologies
have brought the
capacity to
change a
synthetic
speech, image,
or video to our
fingers.
Furthermore,
video and image
frauds are now
so convincing
that it is hard to
distinguish
between false
 and authentic
content with the
naked eye.
(Z., Conduct a
Systematic
CNN, RNN, GAN The
investigati
The study
highlights
ML and DL have achieved high success in many textual,
auditory, medical imaging, and visual recognition
2024); literature review on also the patterns. Concerning the importance of ML/DL in
Heidari (SLR) considers significant recognizing patterns due to its high accuracy, many
A.; important role of ML researchers argued for many solutions for improving
Navimipo parameter and DL pattern recognition performance using ML/DL methods.
ur N.J.; s such as techniques Due to the importance of the required intelligent pattern
Unal M.; accuracy, in pattern recognition of machines needed in image processing and
Mousavi adaptabilit recognition the outstanding role of big data in generating state-of-
A. y across the-art modern and classical approaches to pattern
various recognition, we conducted a thorough SLR about DL
domains approaches for big data pattern recognition. Similarly, we
such as have classified 60 of the most cutting-edge articles put
textual, forward pattern recognition issues into ten categories
auditory, based on the DL/ML method used: CNN, RNN, GAN,
medical Autoencoder (AE), Ensemble Learning (EL),
imaging, Reinforcement Learning (RL), Random Forest (RF), MLP,
and visual LSTM, and hybrid methods.
recognition.
(M.L., A real-time four-
stage model
Scaled-YOLOv4,
SORT, DSFD
model
achieves
The
proposed
The results are compared with other popular state-of-the-
art models in terms of Mean-Average-Precision, frame
2024); an model is rate and loss of values. The DSFD&MobileNetv2
Belahcen accuracy validated on facemask detectors trained on Wider Face and Real Face
e M.; of 56.2% various mask dataset achieves an accuracy of 99.3%. The
Bourenn and a real- indoor/outdo proposed approach is validated on indoor/outdoor public
ane S. time or public images and video sequences such as wider face dataset,
performan datasets, Oxford Town Center dataset and open access sequences.
ce of 32 including the
frames wider face
per dataset and
second. Oxford Town
Center
dataset.
(S.K., Detection of
human
CNN, ReLU. It
achieves
The study
shows how
here is a demand for accurate and scalable systems to
detect human emotional states, as emotion recognition is
2024); emotional states an DL models, needed to understand the mental status of such persons
Suvvari accuracy like the who cannot communicate their emotions, such as
S.; of 98.10% multiple- disabled people, mentally challenged persons, etc.
Kumar M. column CNN Therefore, EEG signals provide a non-invasive method to
network, identify the emotions of these disabled humans. The
effectively research community has recently been very interested in
classify employing electroencephalography (EEG) for emotion
emotions classification since end-users have wearable EEG
using EEG systems that may offer a portable, cheap, and
signals, straightforward technique for identifying emotions. DL
especially models have recently been extensively used to extract
for characteristics and recognize emotions from EEG
 individuals recordings. Apart from that, various papers were
facing reviewed in this research.
communicati
on
difficulties.
(P., Detecting driver
somnolence
3D The study
integrates
The study
underscores
It demonstrates promising advancements in each of
these key considerations, showcasing improved real-time
2024); computer the potential detection capabilities, enhanced accuracy in identifying
Kumar vision, of driver states, scalability to handle diverse driving
R.G.; physiologi integrating scenarios, and effective intervention mechanisms.
Kumar cal computer
M.S.; signals, vision,
Liyakat and physiological
K.K.S.; machine signals, and
Vani M.S. learning ML
algorithms algorithms
to for driver
emphasiz state
e real- detection
time and
detection, intervention
accuracy, systems.
scalability,
and driver
interventi
on
methods.
(A. Assessing
COVID-19
CNN Accuracies
from
The study
shows that
Assessing severity has remained an open and
challenging problem due to a lack of large datasets, the
G. , severity using 3D below iterative large dimensionality of images for which to find weights,
2024); CT scans. 60% at transfer and the compute limitations of modern graphics
first to learning processing units (GPUs). In this paper, a new, iterative
Alag S.;
81.57% effectively application of transfer learning is demonstrated on the
Duncan
with the analyzes understudied field of 3D CT scans for COVID-19 severity
D.
optimizati COVID-19 analysis. This methodology allows for enhanced
ons, severity performance on the MosMed Dataset, which is a small
reaching from 3D CT and challenging dataset containing 1130 images of
similar scans. patients for five levels of COVID-19 severity.
performan Custom
ce to the shallow CNN
state-of- models were
the-art on trained step-
the by-step on
dataset, more
with much challenging
simpler tasks,
setup leading to
procedure improved
s. performance
on the
 MosMed
Dataset.
(V., Enhanced Data
Protection and
RNN The
developed
The study
shows the
In recent years, Intrusion Detection Systems (IDS)
monitor the computer network system by collecting and
2024); Security in Cloud technique effectivenes analyzing data or information by identifying the behavior
Madiajag Computing achieves s of the of the user or predicting the attacks by the automatic
an M.; outstandin proposed response. So, in this paper, the Blockchain-based African
Rafee g BbAB Buffalo (BbAB) scheme with RNN model is proposed for
S.M.; performan scheme in detecting the intrusion by enhancing security.
Sanju P.; ce, with enhancing
Rehman an security in
T.B.; accuracy cloud
Pattanaik of 99.87% computing.
B. and a high
recall of
99.92%.
(A.T., Integrating
Heuristics and
RNNs, LSTM,
CNN
The e-HGA
significant
The research
contributes
The effectiveness of GA can be improved by adjusting
genetic operators and integrating an efficient heuristic.
2024); Fitness Degrees ly reduces to the These enhancements are integrated into the suggested
Mohama task optimization enhanced hybrid genetic algorithm (e-HGA). The e-HGA
d Z.; completio of task begins with an initial population that includes a solution
Mahiddin n time scheduling derived from a heuristic, which serves as a guiding point
N.A.; Nik compared in cloud toward achieving an optimal makespan solution. The
W.N.S.W. to the computing proposed e-HGA was evaluated in this work for two
HGA, by degrees of fitness, which qualified a chromosome and a
consistent introducing gene to be preferred above their other counterpart in a
ly the data population.
outperfor enhanced
ming it hybrid
across genetic
different algorithm (e-
cloudlet HGA), which
and virtual offers
machine improved
configurati performance
ons. in terms of
task
completion
time and
resource
allocation
(G. D. Leveraging
DWSCNN to
WSRS-DNN-BO,
WSRS-CAN.
Accuracy,
showing
The study
demonstrate
The DWSCNN model classifies the outcomes as correct or
incorrect, and to enhances the prediction of web service
, predict the improvem s the composition scalability and reliability, we optimize the
2024); reliability and ents from efficacy of DWSCNN's weight parameters using the Adolescent
scalability of 41.39% to the Identity Search Algorithm (AISA).
Sridhar
BPEL composite 34.16%. proposed
J.;
web services. technique in
Ramanat
enhancing
han R.;
the
Dharmali prediction of
ngam web service
M.R.D. composition
scalability
and
reliability.
(I., Privacy-
Preserving
Deep learning
techniques are
This
approach
This
approach
The rapid expansion of multimedia data has led to a
significant increase in the demand for efficient and
2024); Multimedia employed in prevents ensures both accurate analysis techniques. Deep learning has
Bazai Analysis conjunction with overfitting efficient emerged as a powerful tool for analyzing multimedia
S.U. differential and analysis and data, due to its ability to learn hierarchical
privacy to safeguard the representations of data. However, the users’ privacy in
improve s sensitive protection of crowdsensing and the use of sensitive data in deep
multimedia data from users' learning raises significant privacy concerns. Sensitive
analysis, unauthoriz privacy data may be exposed when collecting the data, while
including tasks ed access rights. training, or even after the trained model has been
such as image or released. This chapter focuses on how to improve
and video disclosure. multimedia analysis by fusing differential privacy with
processing. deep learning.
(A. Y. , In today's digital
age, CNN a
This survey
paper provides a
It's
crucial
2024); subset of DL, are comprehensive to gain
Ansari widely used for examination and a
M.; Fazli various comparison of thoroug
M.; Ejlali computer vision various CNN h
A.; tasks such as architectures, underst
Shafique image highlighting their anding
M.; classification, architectural and
Henkel J. object detection, differences and perform
and image emphasizing a
segmentation. their respective compar
There are advantages, ative
numerous types disadvantages, analysis
of CNNs applications, of these
designed to challenges, and different
meet specific future trends. CNN
needs and types to
requirements, underst
including 1D, 2D, and
and 3D CNNs, as their
well as dilated, strength
grouped, s and
attention, weakne
depthwise sses.
convolutions,
and NAS, among
others. Each
type of CNN has
its unique
structure and
 characteristics,
making it
suitable for
specific tasks.
(T., Lung Cancer
Identification and
CNNs The study
achieved
This
research
Significant progress has been made in the field of
medical image processing in recent years, owing to the
2024); Classification a contributes use of advanced DL techniques such as CNN. The paper
Ratna Using CNNs in remarkabl significantly used CNNs in the work to investigate a novel strategy for
Kumari Medical Imaging e to the identifying and classifying lung cancer, processing on two
K.; Durga accuracy progress of large datasets: the Lung Image Database Consortium
N.; Raju rate of medical Infectious Disease Research Institute (LIDC-IDRI) and the
P.; 0.994 in image Lung Nodule Analysis (LUNA 16). The study began with
Sukhava classifying processing, image preprocessing to improve image quality, which
si V. lung specifically included Bayesian threshold-based Taylor series (BT-
cancer, in the based TS) denoising.
demonstra diagnosis
ting the and
effectiven treatment of
ess of the lung cancer,
proposed through the
approach. application
of deep
learning
methods.
(H. Enhancing
Building Space
The study
utilizes
Experimen
tal results
The study
concludes
In order to solve the problem that the traditional building
space configuration model cannot meet the optimization
H. , Utilization spatialized demonstra that the of building space characteristics, the author proposes the
2024) through particle swarm te that the proposed optimization of building space utilization based on
Spatialized optimization, proposed spatialized spatialized particle swarm optimization. First, to solve the
Particle Swarm which modifies optimizati particle problem of optimal allocation of space units, the PSO
Optimization the PSO on swarm algorithm is modified to encode the space units by
algorithm to algorithm optimization means of character coding. Secondly, the maximum
encode space performs algorithm standardization method is used for data processing, and
units and effectively effectively the factors affecting space utilization are summarized,
optimize building in optimizes the objective function of optimal allocation of
space allocation. optimizing building architectural space is given from three aspects:
building space economic benefits, social benefits and ecological
space utilization. benefits; The experimental results show that: The
utilization. experimental data is based on the utilization of building
space in these three regions in 2015, and the vector map
is divided into 30 m x A grid of 30 m in size, and all
statistical data and spatial data are projected on each
grid cell.
(Zhan In order to solve
the most difficult
This method has
wide applicability
g Q. , problem of the and can be
2024) architectural applied to most
model neural networks
established by which do not
CNN in solving require any
 specific special functional
problems, which modules of the
results in Alexnet network
parameter model.
overflow and
inefficient
training, an
optimization
algorithm for
nonlinear CNN
based on
improved
chameleon
swarm algorithm
is proposed. This
article mainly
introduces the
use of
Chameleon
Swarm
Optimization
(PSO) algorithm
to research the
parameters of
CNN
architecture,
solve them, and
achieve the
optimization of
the optimization
model. Although
the number of
parameters that
need to be set
up in CNN is very
large, this
method can find
better testing
space for Alexnet
samples with 5
different images.
(Ranie Enhancing Flood
Risk
Dense Optical
Flow Field
Enhances
the
Integrating
dense
However, the current methods only analyze static image
features and ignore the valuable motion information that
ri, Management accuracy optical flow may exist in adjacent frames that are captured minutes
Souza of fields into a apart. In addressing this limitation, our approach involves
measuring deep- computing dense optical flow fields from consecutive
, relative learning images taken by a stationary camera and integrating
Nishiji water workflow these representations into a deep-learning workflow. We
levels. enhances evaluated the capacity of both our method and
ma, the accuracy alternative approaches to measure not only the absolute
Krishn of
measuring
water level but also the relative water level.

amac relative
hari, water levels
in urban
& creeks,
Ueya offering a
promising
ma, approach for
flood risk
2024) managemen
t.
(A. G.- The concealment
of improvised
This finding
demonstrates
A. , explosive that the
2024)G devices in designed neural
dustbins aimed networks are
yasi-
at destroying promising
Agyei A.
people and cutting-edge
property is techniques for
causing the mass detecting
removal of explosives
dustbins from concealed in
public places and dustbins to
vehicular public threaten public
transport in safety. To the
cities around the best of our
world. Such knowledge, this
action of dustbin is the first time
removal results that
in littering, convolutional
stench, pests, neural networks
contamination of have been
water bodies, the proposed to
spread of identify
diseases, and explosives
increased concealed in
greenhouse dustbins.
gases. The Graphical
current solutions abstract: (Figure
to the problem presented.)
are blast-
resistant
dustbins which
are bulky and
expensive, and
transparent
dustbins which
display the awful
appearance of
 wastes in public
places. This
article proposes
equipping
dustbins with
artificial
intelligence-
based classifiers
to detect
explosives
concealed in
wastes in public
dustbins to
minimise the risk
to public safety.
(R. Exploring DL
Models for
ScZ, EEG. The study
examines
It
acknowledg
In the traditional approach, the diagnosis of ScZ primarily
relies on the subject’s response and the psychiatrist’s
R. , Schizophrenia multiple es experience, making it highly subjective, prejudiced, and
2024); Diagnosis using deep substantial time-consuming. In recent medical research,
learning challenges incorporating DL into the diagnostic process improves
Sahana EEG Signals
models in both performance by reducing inter-observer variation and
B.C.;
created clinical and providing qualitative and quantitative support for clinical
Bhandari
specificall technical decisions. Compared with other modalities, such as
A.K.
y to aspects. magnetic resonance images (MRI) or computed
diagnose tomography (CT) scans, electroencephalogram (EEG)
schizophre signals give better insights into the underlying neural
nia using mechanisms and brain biomarkers of ScZ.
electroenc
ephalogra
m signals.
(S. L. , Application of
deep learning
grading as
compared to
2024); neural networks other existing
Chanchal (DLNNs) for deep learning
A.K.; Kini histopathological architecture on
J.; image the FPGA board.
Upadhya classification is We have
y G.K. thriving and implemented
implementation and analyzed
of these performances of
networks on DL architectures
edge devices has on a Field
been gaining the Programmable
ground Gate Array
correspondingly (FPGA) board.
due to high Results yield that
computational the Inception-V3
power and low network provides
latency better accuracy
requirements. for kidney cancer
 This paper detection as
designs an compared to
automated other deep
system that learning models
classifies on Kidney
histopathological histopathological
kidney cancer images.
images.
(A. Developing a
Predictive
1D-CNN, MPPP-
SSGSO
Experimen
tal
The study
presents a
Initially, the data are manually collected from the
patients to predict five diseases using 5 G frameworks.
D. , Framework for validation novel Then, the collected data are pre-processed to obtain
2024)D Diagnosing Five
Diseases Using
demonstra
tes that
framework
for disease
high-quality data using the techniques like contrast
enhancement, median filtering, fill empty space, remove
as A.
5G and Deep the diagnosis, repeated value and stemming. The pre-processed data
Learning proposed integrating are taken for extracting the features using a One-
framework deep Dimensional Convolutional Neural Network (1D-CNN) to
outperfor learning obtain the deep features. The parameters like hidden
ms techniques neuron count and epoch are tuned by the proposed
conventio with Modified Predator Presence Probability-based Squirrel
nal ensemble Search-Glowworm Swarm Optimization (MPPP-SSGSO)
technique boosting algorithm to enhance the variance. Then, the extracted
s, models. features acquired using the 1D-CNN are given to the
achieving ensemble boosting-based models for predicting the
an score, which is combined by comprising approaches like
accuracy Adaptive Boosting (AdaBoost), eXtreme Gradient
rate of Boosting (XGBoost) and Category Boosting (CatBoost).
91.34% in
diagnosin
g the five
diseases.
Balasubr Enhancing FOADLM, RNN. The study The study It is challenging to predict Diabetes disease accurately.
amaniya Diabetic Disease attains a proposes a The existing systems face data over-fitting issues,
n D.; Detection remarkabl novel convergence problems, non-converging optimization
Husin e 99.23% approach for complex predictions, and latent and predominant feature
N.A.; accuracy detecting extraction. These issues affect the system's performance
Mustaph in diabetic and reduce diabetic disease detection accuracy. Hence,
a N.; predicting disease the research objective is to create an improved diabetic
Sharef diabetic using a disease detection system using a Flock Optimization
N.M.; Aris disease Flock Algorithm-Based Deep Learning Model (FOADLM) feature
T.N.M. cases, Optimization modeling approach that leverages the PIMA Indian
Algorithm- dataset to predict and classify diabetic disease cases.
Based Deep The collected data is processed by a Gaussian filtering
Learning approach that eliminates irrelevant information, reducing
Model. the overfitting issues. Then flock optimization algorithm
is applied to detect the sequence; this process is used to
reduce the convergence and optimization problems.
Tunio The existing mean absolute
M.H.; Li yield estimating error (MAE),
J.P.; Zeng models exhibit scaled MSE, and
X.; deficiencies in scaled MAE.
Akhtar F.; insufficient These results
Shah examination of emphasize the
S.A.; hyperparameters method's
Ahmed , lack of significance,
A.; Yang robustness, offering valuable
Y.; Heyat restricted insights for crop
M.B.B. transferability of yield estimation,
meta-models, benefiting
and uncertain farmers and the
generalizability agricultural
when applied to sector.
agricultural data.
This study
presents a novel
meta-knowledge-
guided
framework that
leverages three
diverse
agricultural
datasets and
explores meta-
knowledge
transfer in
frequent
hyperparameter
optimization
scenarios.
Heidari With the rapid As the threat of
A.; advancement of deepfake videos
Navimipo DL technology, continues to
ur N.J.; the creation of grow, our
Dag H.; convincingly comprehensive
Talebi S.; realistic solution provides
Unal M. deepfake videos an effective
has become means to protect
increasingly the integrity and
prevalent, raising trustworthiness
serious concerns of visual media,
about the with far-reaching
potential misuse implications for
of such content. both industry
Deepfakes have and society. This
the potential to work stands as a
undermine trust significant step
in visual media, toward
with implications countering the
for fields as deepfake
diverse as menace and
journalism, preserving the
 entertainment, authenticity of
and security. This visual content in
study presents a rapidly
an innovative evolving digital
solution by landscape.
harnessing
blockchain-based
federated
learning (FL) to
address this
issue, focusing
on preserving
data source
anonymity. The
approach
combines the
strengths of
SegCaps and
convolutional
neural network
(CNN) methods
for improved
image feature
extraction,
followed by
capsule network
(CN) training to
enhance
generalization. A
novel data
normalization
technique is
introduced to
tackle data
heterogeneity
stemming from
diverse global
data sources.
Moreover, TL and
preprocessing
methods are
deployed to
elevate DL
performance.
(S. Multimodal
Approaches for
CNNs, RNNs. The
review
Multimodal
approaches
During and after the years of the COVID-19 pandemic,
researchers and domain experts put all their effort into
C. , COVID-19 identifies leveraging the discovery of accurate and reliable techniques for the
2024); Diagnosis multiple deep detection and diagnosis of this disease in potentially sick
multimoda learning patients. In the meanwhile, DL techniques are
Sansone
l solutions techniques continuously improving and expanding, becoming more
C. for COVID- have shown and more efficient and compatible in several fields of
19 promise in study and with different kinds of data. This huge but
diagnosis, enhancing heterogeneous set of data cannot be fully exploited if DL
highlightin COVID-19 models are not designed to be compatible with different
g their diagnosis sources of data at the same time, therefore multimodal
compatibil accuracy by approaches were designed and adopted, resulting in
ity with integrating better prediction results than the classic approaches.
diverse diverse data Given these premises, several multimodal solutions for
data sources. COVID-19 diagnosis were built in these years, but it may
sources result hard to have a complete overview of the current
and their state-of-the-art.
effectiven
ess in
improving
prediction
accuracy
compared
to
traditional
methods.
(H. The study aims
to address data
CNNs, MLP The
experimen
The scheme
defends
The experimental results show that the proposed method
achieved a peak accuracy of 96.6% while scrutinizing the
N. , protection issues tal results smart power original MSU-ORNL data feature set and a maximum
2024); in smart power show that grids against accuracy of 99% when analysing the selected feature
grids by the cyber- set. Therefore, the proposed intrusion classification
Ullah F.;
proposing an proposed attacks by scheme may protect smart power grid systems against
Srivastav
intrusion method combining cyber security attacks.
a G.
classification achieved DL and
scheme capable a peak metaheuristi
of identifying accuracy c
various cyber- of 96.6%. optimization
attacks. , enhancing
security and
operational
reliability.
(D. Developing a
more efficient
NCNN, CNN. The NCNN
model
The study
concludes
Moreover, comparative experiments on the CIFAR-100,
CIFAR-10, MNIST, and ImageNet datasets are conducted
S. , neural network achieves that the to fully demonstrate the effectiveness and feasibility of
2023); structure for competitiv NCNN the proposed NCNN model on image classification. In
image e model, addition, the success of the NCNN also provides new
Chen L.;
classification performan Leveraging ideas for future network structure design. The source
Du X.;
tasks. ce while the code is publicly available at https://github.com/LongJin-
Liu M.;
potentially connectivity lab/Nematode-Connectome-Neural-Network.
Jin L.
reducing architecture
model of C. elegans
complexit offers a
y and promising
computati approach to
onal address the
overhead. challenges
 associated
with
traditional
CNNs in
image
classification
tasks.
(Z. Assessing the
Impact of COVID-
AQI, SAC. Our deep
learning
The deep
learning
These findings demonstrate the effectiveness of strict
lockdown policies in mitigating air pollution in highly
Z. , 19 Lockdowns on analysis analysis affected regions. Moreover, the deep learning framework
2023); Air Quality Index estimates shows enables more reliable AQI forecasts with reduced
(AQI) in China the notable prediction errors of about 47% for Wuhan and 67% for
Wu J.; Cai
lockdown reductions in Shanghai, highlighting the utility of DL techniques in
F.; Zhang
effects on AQI levels understanding and predicting air quality dynamics during
S.; Wang
AQI as during the public health emergencies.
Y.-G.
−25.88 in COVID-19
Wuhan lockdowns,
and with impacts
−20.47 in of -25.88 in
Shanghai. Wuhan and -
20.47 in
Shanghai.
(P. R. , Combating False
Information
Bi-LSTM. The Bi-
LSTM
The study
introduces a
Generally, people share information with their friends
and groups without inspecting the posts’ veracity, which
2024); Propagation in model new method causes false information propagation in the network.
Jain V.; Social Media achieves for detecting Moreover, detecting false news and rumors in such a
Shokeen 99.63% and massive load of unstructured information is a very
J.; Balyan accuracy preventing tedious task. Results, many literature papers explored
A. and a rumor different machine learning and deep learning approaches
remarkabl spread in to detect the presence of rumors on social media
y low false social media networks. Although detection of misleading news and
positive using rumors is not sufficient, therefore, we have proposed a
rate of blockchain model for the detection and prevention of transmitted
0.13%. and deep rumors in this paper.
learning.
(H. Enhancing
Precision and
CNN The
combined
The study
presents a
The research introduces an innovative method that
integrates bispectrum characteristics, deep learning
A. , Efficiency in bispectru new method techniques, and machine learning algorithms to address
2023); Protein Family m and that merges the limitations of traditional methods in protein family
Identification deep bispectrum identification. By numerically representing protein
Al
learning characteristi sequences, employing bispectrum analysis, and utilizing
Fahoum
approach cs with deep various CNN topologies for feature extraction, the
A.; Zyout
markedly learning to method selects robust features for classifying protein
A.;
improves enhance families.
Qasmieh
protein protein
I.A.
family family
identificati identificatio
on, n.
outperfor
ming
 existing
methods
with
superior
classificati
on
metrics.
(M. H. Developing an
Efficient CNN-
CNN The
lightweigh
In
conclusion,
X-ray is a useful imaging modality widely utilized for
diagnosing COVID-19 virus that infected a high number
, Based Model for t model the of people all around the world. The manual examination
2023); Automated proposed lightweight of these X-ray images may cause problems especially
COVID-19 for CNN model when there is lack of medical staff. Usage of deep
Avşar E.
Diagnosis from automate demonstrate learning models is known to be helpful for automated
X-ray Images to d COVID- d superior diagnosis of COVID-19 from the X-ray images. However,
Support Medical 19 performance the widely used convolutional neural network
Staff and diagnosis in classifying architectures typically have many layers causing them to
Enhance achieved lung be computationally expensive. To address these
Diagnostic an diseases, problems, this study aims to design a lightweight
Efficiency accuracy making it differential diagnosis model based on CNN
of well-suited
89.89%, for
surpassing deployment
pre- on devices
trained with
benchmar constrained
k models computation
that al
achieved capabilities.
a
maximum
accuracy
of 85.7%.
(S.K. Enhancing
Neural Network-
The study
employs a hybrid
Experimen
tal results
The study
demonstrate
Several researchers proposed neural network-based drug
discovery techniques to fight against the pandemic;
P. , Based Drug architecture that indicate s the utilizing neural networks has limitations (Exponential
2024); Discovery for integrates neural that effectivenes time complexity, Non-Convergence, Mode Collapse, and
COVID-19 networks with Remdesivi s of the Diminished Gradient). To overcome those difficulties, this
Gupta
Treatment other r and proposed paper proposed a hybrid architecture that will help to
M.K.;
computational Dexameth hybrid repurpose the most appropriate medicines for the
Banerjee
methods for drug asone are architecture treatment of COVID-19. A brief investigation of the
A.; Shai
discovery. the most in sequences has been made to discover the gene density
R.;
effective identifying and noncoding proportion through the next gene
Shivaku
drugs for potential sequencing. The paper tracks the exceptional locales in
mara P.
COVID-19 drugs for the virus DNA sequence as a Drug Target Region (DTR).
treatment, COVID-19
with treatment
effectiven by analyzing
ess rates gene
of 97.41% correlations
and and drug
 97.93%, databases.
respective
ly.
(M. J. , A Fully Pipelined
and Quantized
FPGA, FPQNet,
CNN.
Results
demonstra
The study
introduces
The study focuses on speeding up a type of artificial
intelligence called Convolutional Neural Networks (CNNs)
2023); CNN FPGA te that FPQNet, an used in computer vision tasks. There's a trade-off
Al-Ars Z.; Accelerator for FPQNet FPGA-based between flexibility and speed when placing these CNNs
Hofstee Optimized High achieves CNN on special hardware called FPGAs.
P.; Chang Frame Rate low accelerator
Y.; Zhang Video Processing latencies, optimized
B. with for high
optimized frame rate
designs video
for various processing.
CNN
architectu
res.
(A. P. , IFCnCov:
Integrating IoT,
Deep learning
approaches are
IFCnCov
shows
The
proposed
This study addresses the challenge of remotely
diagnosing coronary diseases amid the COVID-19
2023); Fog Computing, employed in outstandin IFCnCov pandemic, where detection methods face hurdles due to
Parhi M.; Cloud training the two- g system limited data and validation testing. It explores IoT
Pattanay Computing, layered performan demonstrate applications leveraging cloud computing (CC) to enhance
ak B.K.; Ensemble architecture of ce with s significant e-Healthcare systems, despite CC's drawbacks such as
Pati B. Learning, and IFCnCov using high advancemen latency, bandwidth, energy consumption, and security
Deep Learning symptom-based accuracies ts in remote concerns.
for Real-Time and chest X-ray , diagnosis of
Remote imaging precision, COVID-19
Diagnosis of datasets. sensitivity patients,
COVID-19 , offering high
Patients specificity, accuracy,
and F1- precision,
scores in sensitivity,
both and
stages: specificity.
97.71%,
98.64%,
97.38%,
98.52%,
98.19%,
98.92%,
97.21%,
98.34%,
97.79%,
and
98.72%.
(Y. A. , Developing a
Computer-Aided
CNN The CNN-
based
The study's
experimenta
Furthermore, the experimental results of the improved
model were compared to various CNN-based
2023)A Intelligent brain l results architectures using key performance metrics and were
nagun Y. System for Early tumor underscore shown to have a strong impact on tumor categorization.
Brain Tumor diagnosis the
 Diagnosis Using system effectivenes
MRI achieved s of the
outstandin proposed
g metrics: CNN-based
99.85% brain tumor
accuracy, diagnosis
99.89% system.
AUC,
98.16%
precision,
98.17%
recall, and
98.21%
f1-score.
(Ö., Stone formation
in the kidneys is
The observation
is that the
2023); a common Inception-V3
Bilgehan disease, and the model is
B.; high rate of successful in
Kneebon recurrence and detecting kidney
e E.; morbidity of the stones of small
Mirzaei disease worries size. The
O. all patients with performance of
kidney stones. the Inception-V3
There are many Model is at a
imaging options high level and
for diagnosing can be used for
and managing clinical
kidney stone applications. The
disease, and CT research helps
imaging is the the radiologist
preferred identify kidney
method. stones with less
Objectives: computational
Radiologists cost and
need to manually disregards the
analyse large need for many
numbers of CT experts for such
slices to applications.
diagnose kidney
stones, and this
process is
laborious and
time-consuming.
This study used
deep automated
learning (DL)
algorithms to
analyse kidney
stones. The
 primary purpose
of this study is to
classify kidney
stones
accurately from
CT scans using
deep learning
algorithms. The
Inception-V3
model was
selected as a
reference in this
study.
(H. Framework for
Detecting
CNN, Bi-LSTM,
AM.
Results
show
The
proposed
The multi-head attention mechanism outperforms other
models in detecting fake news, particularly when
X. , COVID-19 Fake improvem framework adjusted with various attention mechanisms. There are
2023); News in ents in enhances notable differences in sentence length and topic
Emergencies these fake news distribution between real and fake news. This model
Wang Y.;
metrics of detection serves as a resource and platform for guiding public
Zhang
at least accuracy in opinion during critical emergency news detection.
J.Z.;
1% emergencies
Zheng
compared by
L.J.;
to leveraging
Kamal
baseline outlier
M.M.;
models. knowledge
Arya V.
and
improving
evaluation
metrics.
(F., Brain tumor
segmentation in
Improving
specificity to
2023); medical imaging 0.96 and the dice
Nadeem is a critical task coefficient to
M.; Abrar for diagnosis and 0.89 with the
M.; Amin treatment while increase in
F.; Salam preserving clients from 50
A.; Khan patient data to 100.
S. privacy and Furthermore, the
security. proposed
Traditional approach
centralized outperforms
approaches often existing CNN-
encounter back and RNN-
obstacles in data based methods,
sharing due to achieving higher
privacy accuracy,
regulations and enhanced
security performance,
concerns, and increased
hindering the efficiency. The
 development of findings of this
advanced AI- research
based medical contribute to
imaging advancing the
applications. To field of medical
overcome these image
challenges, this segmentation
study proposes while upholding
the utilization of data privacy and
federated security.
learning. The
proposed
framework
enables
collaborative
learning by
training the
segmentation
model on
distributed data
from multiple
medical
institutions
without sharing
raw data.
(S. Existing security
models either
The modified
PoMT consensus
G. , work on pre- is capable of
2023); emptive selecting
analysis, real- overlapping
Priakanth
time analysis or miner nodes,
P.;
post-processing while
Srivastav
analysis of incorporating
a G.; Joe
different packet load balancing
C.V.
types. These for optimal
operations add energy
to the consumption
complexity of during the
data processing, mining process.
which reduces Due to these
Quality of enhancements,
Service (QoS) the proposed
levels of IoMT model is able to
deployments. To reduce mining
overcome these delay by 8.5%,
issues, this while reducing
paper proposes energy
design of a consumption by
sharded chain 3.2%, and
based high reducing access
 efficiency IoMT delay by 4.9%
security model under real-time
that uses clinical
bioinspired scenarios.
optimizations.
The proposed
model initially
deploys a novel
single chained
Proof-of-Medical-
Trust (PoMT)
based consensus
technique, which
assists in
identification of
high-trust
miners.
(A. A time series is a
sequence of
With respect to
existing
C. , time-ordered literature, the
2023); data, and it is major advantage
generally used to of the work
Capone
describe how a consists in
V.;
phenomenon describing the
Iannuzzo
evolves over most recent
G.;
time. Time series architectures for
Camastr
forecasting, time series
a F.
estimating future forecasting, such
values of time as Graph Neural
series, allows the Networks, Deep
implementation Gaussian
of decision- Processes,
making Generative
strategies. Deep Adversarial
learning, the Networks,
currently leading Diffusion Models,
field of machine and
learning, applied Transformers.
to time series
forecasting can
cope with
complex and
high-dimensional
time series that
cannot be
usually handled
by other
machine learning
techniques.
(K. With the scale
and dimension of
The scalability of
ISO is
W. , engineering investigated on
2023); optimization and CEC2017
real-world (30Dim, 50Dim,
Guo M.;
problems 100Dim) and the
Dai C.; Li
increasing, it will latest CEC2022
Z.
be difficult to (10Dim, 20Dim)
find the optimum suites. The
solutions. This results reveal
paper proposes a that compared to
novel people other
identity competitive
attributes-based algorithms, ISO
heuristic possesses best
technology, computing
named the performance
People Identity with ranking the
Attributes-based first in all
Information- competitors. In
learning Search addition, the
Optimization proposed ISO
(ISO), inspired by and 12
people's competitors
psychological consider 10
assessment and constraint
learning engineering
behaviors for optimization
information problems and
resources based the real
on the social application of
status-based path planning
identity attribute with multiple
and the self- obstacles,
survival suggesting that
demands-based ISO possesses
identity significant
attribute. For the optimization
former, a four- performance.
level information
delivery
mechanism is
constructed
through leader,
manager,
executor, and
freelancer,
emphasizing
global
optimization. For
 the latter, the
staged
transforming
model based on
random
accumulation
and directed
induction
behaviors are
constructed
according to self-
survival
demands,
emphasizing
global
optimization and
exploration.
(J. G. , The intersection
between
By comparing
these findings
2023); technology and with previous
Ocampo education has research, this
Edye E.; taken on study highlights
Revelo- unprecedented the synergistic
Tapia S.; relevance, driven value of these
Villegas- by the promise technologies and
Ch W. of transforming positions itself as
teaching and a critical
learning through resource to guide
advanced digital future
tools. This study developments
proposes a and
comprehensive improvements in
exploration of the education
how cloud sector in a
computing and digitally
artificial advanced world.
intelligence
converge to
impact
education,
focusing on
accessibility,
efficiency, and
quality of
learning. A
mixed-research
design identified
a 25%
improvement in
the
 personalization
of educational
content thanks
to AI and a 60%
(Y.E.I., While reverse
transcription
Nevertheless,
important
2023); polymerase challenges must
Abdoun chain reaction be addressed,
O. (RT-PCR) is the such as limited
standard access to
diagnostic test, it datasets and the
may yield false need for rigorous
negative and model validation.
misleading Additionally, the
results. Artificial generalization of
intelligence (AI) models to
systems are different
accelerating the populations and
transformation of contexts needs
the medical field, to be examined.
particularly in Our findings
early detection underscore the
and diagnosis. need for future
Recent research research
has combined AI directions,
with medical including the
imaging exploration of DL
modalities, such for smaller
as chest X-ray datasets,
(CXR) and CT, to enhancing model
detect the virus, performance for
aiding doctors in complex cases,
making decisions and designing
and reducing practical AI
misdiagnosis systems for
rates. In this deployment in
article, we clinical settings.
conducted a
systematic
review of high-
quality articles
published in
high-impact
journals that
examined CNN-
based methods
for detecting
COVID-19 from
radiographic or
CT images and
 discussed
associated
issues.
(L.M., Developing a
Human Crowd
CNN The
MobileNet
The study
demonstrate
The convolutional neural network model was trained on a
dataset of 9600 images depicting humans, cyclists, and
2023); Detection Model V2 model s the motorcyclists. The dataset included 200 images allocated
Munir R. Using UAVs to achieves feasibility of for testing and hyperparameter tuning. These images
Monitor Social high using UAVs were extracted from videos recorded by UAVs above the
Distancing and precision equipped Institut Teknologi Bandung area, capturing diverse crowd
Mitigate COVID- (84.9%) with human scenarios throughout the day.
19 Spread and recall crowd
(87.8%) in detection
detecting models to
crowds. monitor and
enforce
social
distancing
measures
during the
COVID-19
pandemic.
(A. Unsupervised
Framework Using
HR, HRV. The
unsupervi
The study
shows how
However, many AI solutions rely heavily on supervised
learning techniques, which encounter issues like human
H. , Smartwatch Data sed unsupervise labeling reliability and time-intensive data annotation.
2023); for COVID-19 learning d learning This study introduces an innovative unsupervised
Detection and framework methods framework that utilizes smartwatch data to detect and
Hijazi H.;
Monitoring in in the using monitor COVID-19 infections. Longitudinal data, including
Talib
Fully Vaccinated study smartwatch heart rate (HR), heart rate variability (HRV), and physical
M.A.;
Adolescents and achieves a data can activity measured through step counts, are continuously
Afadar Y.;
Young Adults Silhouette detect and monitored from volunteers.
Nassif
score of monitor
A.B.;
0.55, COVID-19
Nasir Q.
indicating infections.
robust
clustering
performan
ce.
(R.S., Developing an
Efficient Method
DSNN The DSNN
algorithm
The study
introduces a
COVID-19, declared a global pandemic, poses significant
challenges, including pollution caused by fly infestation
2023); for Early COVID- demonstra new method and vulnerabilities in the international economic
Dhanesh 19 Detection tes high for early community. Early detection of pre-malignant diseases
L.; using Deep detection COVID-19 and effective isolation of COVID-19 and non-COVID-19
Aljanabi Learning accuracy detection patients are critical amidst the ongoing coronavirus
M.; Rao Techniques and for COVID- using DL crisis. To address these challenges, a COVID-19
T.V.; Recursive 19, and RRFS recognition system using DL and RRFS is proposed.
Thiruveni Relational providing techniques.
M.; Feature Selection reliable
Mahalaks (RRFS) performan
hmi S. ce
compared
 to
previous
methods.
(E., The study aims
to address the
TRAE, TA. The study
show
The study
presents an
The proposed architectures achieve higher detection
accuracy, surpassing CNN, GAN, and CycleGAN baselines
2023); challenge of substantia innovative by up to 9.0%, 6.5%, and 6.0% at ×8 upscaling factor.
Scarpiniti reliable and l approach to
M.; accurate improvem detect
Momenz automatic ents in COVID-19
adeh A. detection of detection from low-
COVID-19 accuracy resolution
disease from compared and noisy CT
low-resolution to leading scans.
(LR) and noisy CNN/GAN/
CT scans by CycleGAN
proposing a benchmar
novel k
architecture that architectu
combines Single- res for
Image Super- Single
Resolution (SISR) Image
and classification Super-
techniques. Resolution
(SISR).
(A. P. , Maritime vessels
provide a wealth
The XGBoost
model tuned
2023); of data using boosted
Damašev concerning PSO attained an
ičius R.; location, overall accuracy
Jovanovic trajectories, and of 99.72% for the
L.; speed. However, vessel
Toskovic while these data classification
A.; Simic are meticulously problem, while
V.; monitored and the LSTM model
Bacanin logged to attained a mean
N.; maintain course, square error
Zivkovic they can also (MSE) of
M.; provide a wealth 0.000098 for the
Spalević of meta marine trajectory
P. information. This prediction
work explored challenge. A rigid
the potential of statistical
data-driven analysis of the
techniques and classification
applied artificial model was
intelligence (AI) performed to
to tackle two validate
challenges. First, outcomes, and
vessel explainable AI
classification was principles were
 explored through applied to the
the use of determined best-
extreme gradient performing
boosting models, to gain a
(XGboost). better
Second, vessel understanding of
trajectory time the feature
series impacts on
forecasting was model decisions.
tackled through
the use of LSTM
networks.
(S. The Internet of
Things (IoT)
UAV-IFM aims to
improve the
M. , allows farmers to assessment
2023) save money and process so that
effort by keeping smart livestock
tabs on crops, farming may be
mapping out more widely
their land, and adopted and
giving them data offers growth-
to develop supportive help
sensible to farmers.
management Conclusions
strategies for gathered from
their farms. this study's
Surveillance, examination of
disaster the UAV-IFM
management, reveal that these
firefighting, instruments
border patrol, correctly forecast
and courier and verify smart
services employ livestock farming
Unmanned Aerial management
Vehicles (UAVs) within the
that are framework of the
originally created assessment
for the military. procedure. The
The segment experimental
focuses on UAVs analysis of UAV-
in livestock and IFM outperforms
agricultural smart livestock
production.This farming in terms
is achieved via of efficiency
employing ratio,
robots, drones, performance,
remote sensors, accuracy, and
and computer prediction.
imagery in
unison with ever-
 improving in-
Depth Learning
for farming.
Deep learning
(DL) algorithms
find many uses
in the
agricultural
sector, from
identifying plant
diseases to
estimating yields
to detecting
weeds to
forecasting the
weather and
determining how
much water is in
the soil
(D. Diabetic
Retinopathy (DR)
EfficientNetB4 is
the most
D. , is caused as a optimal, efficient
2023); result of and reliable DL
Diabetes Mellitus algorithm in
Biswas
which causes detection of DR,
S.K.;
development of followed by
Bandyop
various retinal InceptionResNet
adhyay
abrasions in the V2, NasNetLarge
S.
human retina. and
These lesions DenseNet169.
cause hindrance EfficientNetB4
in vision and in has achieved a
severe cases, DR training accuracy
can lead to of 99.37% and
blindness. DR is the highest
observed validation
amongst 80% of accuracy of
patients who 79.11%.
have been DenseNet201
diagnosed from has achieved the
prolonged highest training
diabetes for a accuracy of
period of 10– 99.58% and a
15 years. The validation
manual process accuracy of
of periodic DR 76.80% which is
diagnosis and less than the
detection for top-4 best
necessary performing
treatment, is models.
 time consuming
and unreliable
due to
unavailability of
resources and
expert opinion.
(R. IoT in healthcare
systems is
ShCNN is
examined using
M. , currently a viable six metrics,
2023); option for including energy,
providing higher- delay, accuracy,
Om
quality medical sensitivity,
Prakash
care for specificity, and
P. G.;
contemporary e- True Positive
Rajeswar
healthcare. Using Rate (TPR), with
i R.;
an Internet of the maximum
Cristin R.
Things (IoT)– energy of
based smart 0.562 J, the least
healthcare delay of 0.452 s,
system, a the highest
trustworthy accuracy of
breast cancer 91.56%, the
classification higher sensitivity
method called of 96.10%, the
Feedback highest
Artificial Crow specificity of
Search (FACS)– 91.80%, and the
based Shepherd maximum TPR of
Convolutional 99.45%.
Neural Network
(ShCNN) is
developed in this
research. To
choose the best
routes, the
secure routing
operation is first
carried out using
the
recommended
FACS while
taking fitness
measures such
as distance,
energy, link
quality, and
latency into
account. Then,
by merging the
Crow Search
 Algorithm (CSA)
and Feedback
Artificial Tree,
the produced
FACS is put into
practice (FAT).
After the
completion of
routing phase,
the breast
cancer
categorization
process is
started at the
base station.
(B., The outbreak of
COVID-19 has
The latest
detection
2023); caused great strategy based
Ikbal havoc and on
A.M.A.; affected many nanotechnology,
Farooqui parts of the biosensor, is said
A.; Sahu world. It has to produce
R.K.; Ruhi imposed a great satisfactory
S.; Syed challenge to the results in
A.; medical and recognizing
Miatmok health fraternity SARS-CoV-2
o A.; with its ability to virus. It also
Khan D.; continue highlights the
Khan J. mutating and successes in the
increasing the research and
transmission development of
rate. Some COVID-19
challenges treatments and
include the vaccines that are
availability of already in use. In
current addition, there
knowledge of are a number of
active drugs nanovaccines
against the virus, and
mode of delivery nanomedicines
of the currently in
medicaments, its clinical trials that
diagnosis, which have the
are relatively potential to
limited and do target COVID-19.
not suffice for
further
prognosis. One
recently
developed drug
 delivery system
called
nanoparticles is
currently being
utilized in
combating
COVID-19.
(C.-H., Economic
forecasting is
The GRU model
demonstrates
2023); crucial in superior
Molefyan determining a performance
e T.; Lin country’s compared to all
Y.-D. economic growth other models in
or decline. terms of MAE,
Productivity and RMSE, and MAPE
the labor force (with an average
must be MAPE of 2.774%)
increased to when forecasting
achieve government
economic spending using
growth, which data from the
leads to the world’s 15
growth of gross largest
domestic product economies from
(GDP) and 1990 to 2020.
income. Machine The results
learning has indicate that the
been used to GRU can be used
provide accurate to provide
economic accurate
forecasts, which economic
are essential to forecasts.
sound economic
policy. This study
formulated a
GRU neural
network model
to predict
government
expenditure, an
essential
component of
gross domestic
product.
(Azee Artificial neural
networks (ANNs)
The results show
that the ConXNet
m, ability to learn, model performed
2023); correct errors, significantly well
and transform a for a relatively
Javaid S.;
large amount of large dataset
Khalil raw data into compared with
R.A.; beneficial the
Fahim H.; medical aforementioned
Althobaiti decisions for models.
T.; treatment and Moreover, the
Alsharif care has ConXNet model
N.; increased in reduces the time
Saeed N. popularity for complexity by
enhanced using dropout
patient safety layers and batch
and quality of normalization
care. Therefore, techniques.
this paper Finally, we
reviews the highlight future
critical role of research
ANNs in directions and
providing challenges, such
valuable insights as the
for patients’ complexity of the
healthcare algorithms,
decisions and insufficient
efficient disease available data,
diagnosis. We privacy and
study different security, and
types of ANNs in integration of
the existing biosensing with
literature that ANNs. These
advance ANNs’ research
adaptation for directions require
complex considerable
applications. attention for
Specifically, we improving the
investigate scope of ANNs
ANNs’ advances for medical
for predicting diagnostic and
viral, cancer, treatment
skin, and COVID- applications.
19 diseases.
(D. Developing a
Privacy-
Chest X-ray
(CXR.
CoviDetec
tor
The study
shows that
Additionally, we generate heatmaps utilising Grad-CAM
and overlay them on the CXR images to present the
C. , Preserving classifies CoviDetecto highlighted areas that were deciding factors in detecting
2023); COVID-19 individuals r, utilizing a COVID-19. Finally, we developed an Android app to offer
Detection Model into three semi- a user-friendly interface.
Das A.;
Using Chest X- categories supervised
Dey A.;
ray Images : COVID- approach,
Banerjee
19 substantially
S.; Golec
positive, enhances
M.;
Viral the accuracy
Kollias
pneumoni of COVID-19
D.;
a, and detection
Kumar Normal, while
M.; Kaur achieving reducing the
G.; Kaur over 99% need for
R.; Arya accuracy extensive
R.C.; across training
Wander four data.
G.; distinct
Wander datasets.
P.;
Wander
G.S.;
Parlikad
A.K.; Gill
S.S.;
Uhlig S.
(E. Although the
disease appears
Third, a
comprehensive
G. , to have lost its comparison is
2023); impact, millions made by
of people have contrasting the
Kaya Y.
been affected by similar and
this virus, and different aspects
new infections of the studies.
still occur. Finally, the
Identifying results, gaps,
COVID-19 and limitations
requires a are summarized
reverse to stimulate the
transcription- improvement of
polymerase COVID-19
chain reaction detection
test (RT-PCR) or methods, and
analysis of the study
medical data. concludes by
Due to the high listing some
cost and time future research
required to scan directions for
and analyze COVID-19
medical data, classification.
researchers are
focusing on
using automated
computer-aided
methods.
(H.M., Development
and Evaluation of
1D-CNN, BiLSTM,
and GRU DL
ML with
TDA
The
proposed
When concerning the UCI-HAR dataset, the best-reported
scores are accuracy and WSM of 100% and 100%,
2023); ML and DL techniques and achieved approach respectively. For the second category experiments, the
Hassan Approaches for evaluated three 95.34% utilizing DL best-reported scores concerning the WISDM dataset are
A.E.-S. Human Activity experiment accuracy techniques accuracy and WSM of 95.34% and 89.62%, respectively.
Recognition categories. on WISDM (1D-CNN, When concerning the UCI-HAR dataset, the best-reported
 Using WISDM and BiLSTM, and scores are accuracy and WSM of 96.70% and 92.57%,
and UCI-HAR 96.70% GRU) respectively. For the third category experiments, the
Datasets accuracy significantly best-reported scores concerning the WISDM dataset are
on UCI- outperforms accuracy and WSM of 99.90% and 99.76%, respectively.
HAR, with traditional When concerning the UCI-HAR dataset, the best-reported
WSM of ML scores are accuracy and WSM of 100% and 100%,
89.62% on approaches respectively.
WISDM with both
and traditional
92.57% on feature
UCI-HAR. extraction
and TDA.
(G. Developing a
Novel High-
Self-supervised
learning
The
method
The
proposed
However, the existing methods usually use supervised
transfer learning from natural images as a pretraining
L. , Accuracy COVID- pretrains models maintains method process. These methods do not consider the unique
2023); 19 Detection on CXR images high outperforms features of COVID-19 and the similar features between
Method Using without labels, detection other state- COVID-19 and other pneumonia. Aim: In this paper, we
Togo R.;
Chest X-ray learning distinct accuracy of-the-art want to design a novel high-accuracy COVID-19 detection
Ogawa
Images features. Batch even with COVID-19 method that uses CXR images, which can consider the
T.;
Considering knowledge a detection unique features of COVID-19 and the similar features
Haseyam
Unique Features ensembling fine- significant methods in between COVID-19 and other pneumonia.
a M.
and Similarities tunes by reduction various
with Other leveraging image in settings,
Pneumonia category annotated reducing the
similarities CXR workload of
within batches, training healthcare
enhancing images providers
detection (e.g., and
performance. using only radiologists.
10% of
the
original
dataset).
(M. C. Validation of
RegNetX032 CNN
RegNetX032
model.
The model
achieved
The robust
deep
In addition, an explainability model was developed that
demonstrated the high performance and high
, Model for COVID- high learning generalization degree of the proposed model in detecting
2023); 19 Detection accuracy model and highlighting the signs of the disease. The fine-tuned
from Chest X-ray and AUC demonstrate RegNetX032 model achieved an overall accuracy score of
Akhloufi
Images Using RT- scores d excellent 96.0%, with an AUC score of 99.1%. The model showed a
M.A.;
PCR as overall performance superior sensitivity of 98.0% in detecting signs from CXR
Bouattan
Reference and in in detecting images of COVID-19 patients, and a specificity of 93.0%
e E.M.;
specific COVID-19 in detecting healthy CXR images.
Abdulnou
scenarios, from chest
r J.; Roux
with X-rays,
S.;
excellent achieving
Bernard
sensitivity high
C.D.
and accuracy,
specificity, sensitivity,
demonstra and
ting specificity.
 robust
performan
ce in
detecting
COVID-19
from chest
X-rays.
(I. A. , Development of
an Automated
Faster-RCNN,
YOLOv3, SSD.
The
experimen
The research
highlights
However, collecting, measuring, and analyzing large
volumes of health-related data like images is a laborious
2023); Deep Learning tal the and time-consuming job for medical experts. Thus, in this
Chehri System for Lung outcomes effectivenes regard, artificial intelligence applications (including
A.; Jeon Nodule demonstra s of deep machine and deep learning systems) help in the early
G.; Classification ted a learning diagnosis of various contagious/ cancerous diseases such
Piccialli F. and Detection in decrease models like as lung cancer.
CT Images in false Faster R-
positive CNN,
occurrenc YOLOv3, and
es and SSD in
enhanced precisely
overall categorizing
accuracy. and
detecting
lung nodules
in CT scans.
Chan Investigating the The study Post- The study However, little is known about how the underground
I.Y.S.; Impact of primarily outbreak, suggests environment, users’ health, and their associations were
Chen H. Underground employs users that despite impacted by the outbreak of the pandemic. In this study,
Environments on multigroup showed the we investigated and compared the impacts of the
Physical and structural increased challenges general underground environment on regular users’
Psychological equation satisfactio of physical and psychological health before and after the
Health modeling to n with underground pandemic. To achieve this aim, the data from 525
analyze the data undergrou spaces, surveys were collected from eleven underground sites,
collected from nd users' followed by an objective field measurement study
surveys and environme satisfaction conducted at five underground sites in Hong Kong pre-
objective field nt factors, and health and post-outbreak of the pandemic.
measurements. along with improved
significant post-
improvem outbreak,
ents in with visual
physical comfort
and playing a
psychologi critical role.
cal health,
with visual
comfort
playing a
crucial
role.
(Gurc Enhancing
Diabetic
InceptionV3, is
proposed for
The hybrid
approach
The study
demonstrate
The International Diabetes Federation (IDF) highlights
diabetes as a rapidly increasing global health concern,
 Retinopathy feature demonstra s the affecting approximately 463 million adults aged 20-79,
Diagnosis extraction from ted its efficacy of with millions more undiagnosed. By 2030, this number is
Performance retinal images, effectiven integrating projected to rise to about 578 million.
Using a Hybrid followed by ess by deep
Model feature selection achieving learning-
using Simulated a 92.55% based
Annealing, and accuracy feature
finally, the in binary extraction
XGBoost diabetic with
algorithm for retinopath metaheuristi
classification. y (DR) c feature
classificati selection
on with methods for
the diagnosing
Messidor-2 diabetic
dataset retinopathy,
using the offering a
XGBoost valuable tool
model. to aid
ophthalmolo
gists in
diagnosis.
(I. A. One of the most
dangerous
and
histogram charac
E.-F., malignancies in teristics, to
2023); global conduct
diseases is skin thorough and
Ali A.M.;
cancer. The best impartial
El-Shafai
way to survive comparisons and
W.; Taha
skin cancer is to identify the best
T.E.; Abd
detect it in its models
El-Samie
early stages. Its according to
F.E.
diagnosis is performance
made by medical training time,
imaging with and storage
dermoscopy, space. By
which provides obtaining an
High-Resolution accuracy of
(HR) lesion 98.9958%, the
images to results show that
simplify the proposed
the detection pro model
cess. outperforms
Dermoscopy is traditional and
not available to previous models
everyone, and for color and
other available texture
imaging reconstruction
technologies, and recognition.
such as mobile
 devices,
deteriorate the
image quality
required for
diagnosis. This
research
presents a
proposed novel
Enhanced Deep
Super-Resolution
Generative
Adversarial
Network (EDSR-
GAN) model to
generate HR skin
disease images
from Low-
Resolution
(LR) ones.
(S. Developing an
Integrated Deep
The study
employs transfer
The
proposed
The study
demonstrate
The COVID-19 virus has fatal effect on lung function and
due to its rapidity the early detection is necessary at the
R. , Learning Model learning to integrated s the moment. The radiographic images have already been
2023); for Early COVID- extract features deep effectivenes used by the researchers for the early diagnosis of COVID-
19 Detection from chest X-ray learning s of 19. Though several existing research exhibited very good
Das A.K.
Using Chest X- and CT images, model combining performance with either x-ray or computer tomography
ray and CT then utilizes achieves chest x-ray (CT) images, to the best of our knowledge no such work
Images capsule networks high and CT has reported the assembled performance of both x-ray
to classify the accuracy images with and CT images.
subset into rates of deep
COVID-positive 98.2% and learning
and normal 97.8% techniques
patients. with x-ray for the early
and CT detection of
images, COVID-19.
respective
ly, and an
average
accuracy
of 98%.
(M. B. In information
theory - as well
The potential for
wide-spread
, as in the applicability is
2023); adjacent fields of indicated, too; in
statistics, particular, we
Stummer
machine deliver many
W.
learning, artificial recent
intelligence, references for
signal processing uses of the
and pattern involved
recognition - distances/diverg
 many ences and
flexibilizations of entropies in
the omnipresent various different
Kullback-Leibler research fields
information (which may also
distance (relative serve as an
entropy) and of interdisciplinary
the closely interface).
related Shannon
entropy have
become
frequently used
tools. To tackle
corresponding
constrained
minimization
(respectively
maximization)
(X. As a branch of
the Internet of
The
experimental
D. , Things (IoT) results show that
2023); dedicated to the classification
medical care accuracy of the
Zhang Y.;
industry, the accident
Li J.; Mao
Internet of diagnosis results
B.; Guo
Medical Things is effectively
Y.; Li G.
(IoMT) includes improved.
networking Compared with
devices and other algorithms,
applications for this model with
medical and multi-mode data
health fusion CNN
information algorithm
technology improves the
applications. The data
IoMT still has transmission
many rate, reduces the
shortcomings, average data
such as unstable delay, and
information improves the
transmission, low real-time
accuracy of fault anomaly
diagnosis and monitoring
classification, capability of
and lack of IoMT without
anomaly increasing the
detection signal leakage
capability. rate, thus further
Therefore, this improving the
paper first overall stability
 comprehensively of IoMT.
describes some
relatively new
research and
popular
background
technologies of
the Industrial
Internet of
Things (IIoT) and
IoMT, and then,
according to the
functional
(R. Subspace
clustering is an
K. , extension of
2023); conventional
clustering
Das P.;
techniques that
Namasud
seeks to find
ra S.
clusters in
different
subspaces within
a data set with
higher
dimensions. But,
recent research
results indicate
that there are
few challenges
to be addressed
efficiently, such
as finding
number of
subspace
clusters
automatically
and getting
optimal
subspaces by
satisfying
multiple
objective criteria.
This paper
focused on these
two challenges
and proposed a
framework called
Subspace
Clustering using
 Evolutionary
algorithm, Off-
Spring
generation and
Multi-Objective
Optimization
(SCEOMOO) to
find the optimal
subspace
clusters.
(A. The profitability
of a supply chain
Multiple
experiments are
S. , (SC) is carried out to
2023); proportional to demonstrate the
the stability of all effectiveness of
Chakrabo
its stakeholders our approach,
rtty R.K.;
as well as their and the
Essam
consistent simulation
D.L.
information results clearly
sharing with an demonstrate the
effective and effectiveness of
efficient information
communication sharing in a
mechanism. supply chain via
Various blockchain, as
inefficiencies, well as that trust
such as the between
bullwhip effect partners tends to
(BWE) and increase overall
product SC efficiency and
unavailability, reduce BWE.
may be caused
by a lack of
coordination in
an SC. The
importance of
sharing
consumer
demand has
been quantified
by
comprehensive
studies under
the assumption
that all SC
participants will
access the same
information.
However, only a
few studies have
 studied the
effect of minimal
coordination or
limited visibility
of information
while considering
their effect on
the overall
efficiency of an
SC.
(A. The Internet of
Things (IoT) is a
On the other
hand, some
N. , communication quality attributes
2023); paradigm and a such as
collection of confidentiality,
Jafari
heterogeneous accessibility, and
Navimipo
interconnected sustainability are
ur N.;
devices. It less considered.
Hosseinz
produces large- Also, none of the
adeh M.;
scale distributed, articles use key-
Sharifi A.
and diverse data value databases
called big data. for data storage.
Big Data This study can
Management help researchers
(BDM) in IoT is develop more
used for effective BDM in
knowledge IoT methods in a
discovery and complex
intelligent environment.
decision-making
and is one of the
most significant
research
challenges today.
There are several
mechanisms and
technologies for
BDM in IoT. This
paper aims to
study the
important
mechanisms in
this area
systematically.
This paper
studies articles
published
between 2016
and August
2022.
 Initially, 751
articles were ide
ntified, but a
paper
(Y. K. , Developing a
Deep Transfer
Deep transfer
learning with
The
models
The
proposed
Once it is vital to diagnose such a disease in a timely
manner, the researchers have focused on computer-
2023); Learning novel fine-tuning achieved method aided diagnosis systems. Chest X-rays have helped
Gürsoy Approach with mechanisms. average efficiently monitor various lung diseases consisting COVID-19. In
E. Novel Fine-Tuning accuracy and this study, we proposed a deep transfer learning
Mechanisms for rates of accurately approach with novel fine-tuning mechanisms to classify
Accurate COVID- 95.62%, identifies COVID-19 from chest X-ray images.
19 Classification 96.10%, COVID-19
from Chest X-ray and from chest
Images 97.61% X-rays,
for 3-class proving its
cases with potential for
fivefold early
cross- diagnosis
validation. and
monitoring.
(J. X. , Battlefield
information is
Furthermore, a
target maneuver
2023); generally tendency
Chen M.; incomplete, function is
Fang W. uncertain, or proposed to
deceptive. To screen out the
realize enemy possible
intention deceptive attack
recognition in an intention. The
uncertain and one-to-one air
incomplete air combat
combat simulation shows
information that the
environment, a proposed
novel intention method has
recognition advantages in
method is both accuracy
proposed. After and efficiency of
repairing the state prediction
missing state and intention
data of an recognition, and
enemy fighter, is suitable for
the gated enemy fighter
recurrent unit intention
(GRU) network, recognition in
supplemented by small air combat
the highest situations.
frequency
method (HFM), is
used to predict
 the future state
of enemy fighter.
(A. Identifying
Human Epithelial
Whereas,
average scores
A. , Type 2 (HEp-2) of 86.986%
2023); mitotic cells is a recall, 85.282%
crucial procedure precision, and
Al-antari
in anti-nuclear 78.506% mAP
M.A.;
antibodies are obtained
Hussain
(ANAs) testing, using the Faster
J.; Abdel
which is the R-CNN predictor.
Samee
standard Employing the
N.;
protocol for DAL method over
Alabdulh
detecting four labeling
afith M.;
connective tissue rounds
Gofuku
diseases (CTD). effectively
A.
Due to the low enhances the
throughput and accuracy of the
labor-subjectivity data annotation,
of the ANAs’ and hence,
manual improves the
screening test, prediction
there is a need performance.
to develop a The proposed
reliable HEp-2 framework could
computer-aided be practically
diagnosis (CAD) applicable to
system. The support medical
automatic personnel in
detection of making rapid
mitotic cells from and accurate
the microscopic decisions about
HEp-2 specimen the mitotic cells’
images is an existence.
essential step to
support the
diagnosis
process and
enhance the
throughput of
this test.
(W.N., Developing an
AUTO-HAR
NAS, GA, CNN The
proposed
The study
showcases
In addition, the proposed search space provides a
reasonable degree of depth because it does not limit the
2023); Architecture method the maximum length of the devised task architecture. To test
Alsalama Using Neural achieves effectivenes the effectiveness of the proposed framework for HAR
h H.A.; Architecture high s of the tasks, three datasets were utilized: UCI-HAR,
Hassan Search with accuracy AUTO-HAR Opportunity, and DAPHNET. Based on the results of this
M.M.; Genetic rates of framework, study, it has been found that the proposed method can
Mohame Optimization for 98.5% . utilizing NAS efficiently recognize human activity with an average
d E. Human Activity with GA to accuracy of 98.5% (∓1.1), 98.3%, and 99.14% (∓0.8) for
 Recognition automaticall UCI-HAR, Opportunity, and DAPHNET, respectively.
y create
optimal CNN
architecture
s for human
activity
recognition
tasks.
(R. In order to
provide real-time
Enhanced
optimization
K. , identification of methods,
2023); objects in adaptive
Internet of safeguards, and
Pundir S.;
Things (IoT) additional
Shrivasta
software for the scalability
va A.;
healthcare investigations
Kumar
industry, this should all be the
C.P.; Rao
research focus of future
A.L.N.;
examines the research. Future
Sankhya
incorporation of research around
n A.
cutting-edge real-time
computing with analytics gives
convolutional an intriguing
neural networks opportunity to go
(CNNs). A beyond
deductive recognizing
method is used objects as well.
to methodically Overall, this
optimize CNN research
representations revolutionizes
for constrained real-time object
by resources identification in
edge devices, medical IoT for
and it is based better patient
on an care by
interpretivist advancing the
mindset. To give practical use of
a thorough edge
explanation of technologies and
the technical CNNs.
complexities
concerned, a
design based on
descriptive
statistics is used.
The analysis is
built on
secondary data
gathered from
reliable sources.
 The topics
covered range
from enhancing
CNN machine
learning models
on edge devices
to putting secure
processing of
information
(Y.M., Real-time data
processing and
The study differs
greatly from
2023); distributed other studies in
Eken S. messaging are terms of using
problems that distributed
have been computing and
worked on for a big data
long time. As the technologies on
amount of spatial data to
spatial data combat COVID-
being produced 19. It is
has increased, concluded that
coupled with Pulsar is
increasingly designed to
complex handle large
software amounts of long-
solutions being term on disk
developed, there persistence.
is a need for
platforms that
address these
needs. In this
paper, we
present a
distributed and
light streaming
system for
combating
pandemics and
give a case study
on spatial
analysis of the
COVID-19 geo-
tagged Twitter
dataset. In this
system, three of
the major
components are
the translation of
tweets matching
with user-defined
 bounding boxes,
(Y. C. , COVID-19 has
continuously
given to evaluate
the performance
2023); influenced of the proposed
Fu Z. energy security model, in
and caused an comparison with
enormous impact 4 deep learning
on human life models and 6
and social popular machine
activities due to learning models
the stay-at-home based on 12
orders. After the evaluation
Omicron wave, metrics. Results
the economy and show that CNN-
the energy BiLSTM
system are outperforms all
gradually other models in
recovering, but four cases, with
uncertainty MAPEs ranging
remains due to from 4.0034% to
the virus 5.4774%,
mutations that improved from
could arise. 0.1252% to
Accurate 49.1410%,
forecasting of compared with
the energy other models,
consumed by the which is also
residential and about 5 times
commercial lower than that
sectors is of the CNN and
challenging for 5.9559% lower
efficient than the BiLSTM
emergency on average. It is
management evident that the
and policy- proposed CNN-
making. Affected BiLSTM has
by geographical improved the
location and prediction
long-term accuracy of the
evolution, the CNN and BiLSTM
time series of and has great
the energy potential in
consumed by the forecasting the
residential and energy
commercial consumed by the
sectors has residential and
prominent commercial
temporal and sectors.
spatial
characteristics.
(P.C., Because of its
high incidence
We also found
that the best ML
2023); and high model to use
Nithyapri mortality rate, with the
ya S.; breast cancer ResNet50V2
Sharma requires reliable architecture is
N.; means of early the Light
Maheswa detection and Boosting
ri S.; diagnosis to Classifier (LGB).
Jayaram improve survival This study's
B.; Katale rates for its findings provide
T. patients. With substantial
advantages contributions to
including the field of
increased breast cancer
accuracy and diagnosis thanks
efficiency, DL to its novel
and ML methods approach, in-
are quickly depth
becoming performance
indispensable in analysis, and
the fight against trustworthy
breast cancer. evaluation. In
However, there addition, it may
is a need for help doctors
additional study make better
because existing decisions,
approaches have provide better
scalability and care, and boost
performance results for people
constraints. with breast
Utilizing a pre- cancer.
trained
ResNet50V2 and
ensemble-based
ML approaches,
we offer a hybrid
approach to
reliable breast
cancer diagnosis
in this study.
(W. X. Proposing a
COVID-19
CNN, DCGAN Extensive
evaluation
The study
highlights
Severe COVID-19 has been a global pandemic which
provokes massive devastation to the society, economy,
, Detection s the and culture since January 2020. The pandemic
2023); Scheme Using demonstra effectivenes demonstrates the inefficiency of superannuated manual
MobileNet for te the s of utilizing detection approaches and inspires novel approaches that
Chen B.;
Chest X-ray efficiency edge detect COVID-19 by classifying chest x-ray (CXR) images
Shi H.;
Classification in and computing with deep learning technology. Although a wide range of
Tian H.;
Edge Computing accuracy and researches about bran-new COVID-19 detection methods
Xu X.
to Enhance of the lightweight that classify CXR images with centralized CNN models
 Efficiency and proposed CNN models have been proposed, the latency, privacy, and cost of
Reduce scheme like information transmission between the data resources and
Centralized Data for MobileNet the centralized data center will make the detection
Center Load detecting for COVID- inefficient.
COVID-19 19 detection
over CXR through CXR
images in image
edge classification
computing .
.
(S. L. , Enhancing
Efficiency,
CNNs, SLAM. The
proposed
The study
demonstrate
Detecting protective measures (e.g., masks, goggles and
protective clothing) is a momentous step in the fight
2023); Economy, and WD-PL- s the against COVID-19. The detection mode of unmanned
Zhu W.; Safety in COVID- VINS effectivenes devices based on Simultaneous localization and mapping
Chakrabo 19 Protective method s of the (SLAM) and fusion technology is more efficient,
rty C.; Du Measure was proposed economical and safe than the traditional manual
J.; Yu K. Detection Using evaluated method in detection. In this paper, a tightly-coupled nonlinear
Unmanned using enhancing optimization approach is used to augment the visual
Devices with publicly the accuracy feature extraction of SLAM by the gyroscope of the IMU
SLAM and Fusion available of detecting to obtain a high-precision visual inertial system for joint
Technology EuRoC protective position and pose estimation.
and TUM measures
VI against
datasets COVID-19
and tested using
in lab unmanned
settings devices
with an equipped
unmanned with SLAM
vehicle on and fusion
the technology.
NVIDIA
Jetson-TX2
developm
ent board.
(F. L. , The growing
turbulence of the
The results of the
analysis highlight
2023)L external how the data-
oia F. environment has driven approach
progressively led supports urban
to the necessity decision-making
by organizations process and shed
of exploiting new light on the co-
opportunities evolutionary
provided by perspective as
data-driven heuristic device
approaches for to map the
supporting the interactions
even more settled in the
complex decision networks
 making between local
processes. The governments,
new digital data-driven
environment has technologies,
led to the and citizens. In
development this sense, this
and adoption of chapter offers
innovative interesting
approaches; also insights,
in the urban potentially
context which capable of
has always been generating
characterized by useful
different, implications for
interconnected, both researchers
and dynamic and professionals
dimensions. in the public
Urban sector.
governance
models have
been enhanced
by smart
technologies,
Esfandiar The Bat-inspired Experimental
i A.; algorithm (BA), results show that
Khalooza as one of the FOBA performs
deh H.; swarm better than all
Farivar F. intelligence compared
algorithms, has a algorithms.
high potential for Furthermore, the
solving global proposed
optimization algorithm is used
problems. This to optimize the
algorithm weights and bias
possesses an of the MLP neural
additional network using
inherent six classification
capability datasets. The
compared to results
other swarm demonstrate the
intelligence effectiveness
algorithms, and efficiency of
which is the FOBA.
inclusion of a
local search
mechanism.
Although this
increases the
convergence
speed of the
 algorithm,
excessive focus
on exploitation,
especially in the
initial iterations,
may lead to
premature
convergence and
stagnation at the
local optimum. In
this paper, to
address this
drawback and
strengthen the
exploration
capability while
also achieving a
balance between
it and the
exploitation
capability, an
improved version
of BA, called
fractional-order
BA (FOBA), is
introduced.
Nessipkh Enhancing CNN model The Deep The study However, while celebrating these outcomes, the study
anov D.; Accuracy in optimized for CNN underscores also emphasizes the importance of addressing potential
Davletov Lower distinguishing model the hurdles in clinical application, including ethical
a V.; Respiratory Tract pneumonia from achieved transformati considerations, model scalability, and adaptability to
Kurmanb Infection healthy cases in impressiv ve capability evolving pulmonary disease profiles.
ekkyzy Diagnosis Using chest X-ray e of Deep CNN
N.; Deep CNNs: A images. performan in
Omarov Chest X-ray ce pneumonia
B. Study metrics, diagnosis via
with an X-ray
accuracy imaging.
score of
98.1%.
Mishra S. More and more Therefore, it may
studies are be possible to
looking at the construct a small
use of dynamic and efficient
off-chain pupil DRL model
transportation in and utilise it to
IoT systems that optimise the
are based on long-term
payment channel efficacy of
networks (PCNs). operations in off-
Flexible routing chain networks
 in PCN-based on limited-
IoT, where resource IoT
devices often devices. Results
have little from computer
resources, has simulations show
two key that the
challenges, suggested DRL
however. The algorithm
primary problem improves PCN
is figuring out transaction
how to make performance
PCNs with compared to the
variable channel baseline
capacity work techniques, while
efficiently for using just 10% of
long-term the resources
transactions. The available in the
second issue is original
how to maximise instructor model.
transaction
efficiency (the
number of
successful
payments and
the proportion of
those that go
through) by
employing a
lightweight
routing strategy
on IoT gadgets.
Uwiragiy Predicting These findings
e B.; building energy underscore the
Duhirwe consumption is potential of the
P.N.; Seo important for adopted model
H.; Yun energy efficiency in addressing
G.Y. and reducing data scarcity in
carbon existing and new
emissions. buildings for
However, deep accurate cooling
learning (DL) energy
models for prediction.
energy
consumption
forecasting often
have limited
predictive
generalization
and lack
explainability. To
 address these
challenges, this
research adopts
a sequential
attention deep
learning
architecture
(SADLA) that
uses attention
mechanisms to
learn the
importance of
different features
for predicting
cooling energy
consumption,
which helps
building
occupants and
managers to
make informed
decisions about
energy
optimization. The
model was
trained and
tested on
comprehensive
and scarce
datasets in terms
of recording
periods and
predictor
numbers.
Zhang Y.; A major concern s-PBIL could
Chen C.; in designing accurately learn
Wu T.; sensor networks the required
Miao C.; is the information, and
Ding S. deployment automatically
problem. learn a sparse
However, distribution.
establishing an Third, a mixed-
efficient variable
algorithm for the surrogate with
real-world unifying space
deployment under Bayesian
problem is model
challenging due management is
to three issues, incorporated to
which are 1) the reduce the
 realistic mixed- expensive
integer nonlinear computational
programming cost. Experiment
problem (MINLP) results on real-
with mixed- world
variable; 2) the deployment
combinatorial scenarios scaling
subset selection from small-size
problem; and 3) to large-size
the expensive show the
computational effectiveness of
cost for fitness the proposed
evaluation in the algorithm.
3-D coverage
problem.
Sanjalaw Numerous detects
e Y.K.; Al- abnormal abnormal attacks
E'mari transactions within the
S.R. have been Ethereum
exposed as a network. ATD-
result of targeted SGAN leverages
attacks on a semi-
Ethereum, such supervised
as the Ethereum generative
Decentralized adversarial
Autonomous network for this
Organization purpose. The
attack. Exploiting results
vulnerabilities in demonstrate that
smart contracts, ATD-SGAN
malicious users significantly
can pursue their enhances the
own illicit performance of
objectives state-of-the-art
through IDSs. It achieves
abnormal an increase in
transactions. detection
Consequently, accuracy from
identifying these 3.78% to 11.05%
malevolent and reduces the
users, implicated false alarm rate
in fraudulent from 42.29% to
activities and 0.15%.
their attribution, Moreover, ATD-
becomes SGAN notably
exceedingly improves the F1-
complex. measure,
Cryptocurrency ranging from
transactions 10.39% to
used for 3.79%,
 malicious compared to the
purposes, current IDSs.
employing
Arul Sample-Efficient The study The model The study COVID19 is an infectious disease that is rapidly producing
Sowmiya Deep Learning utilizes a deep achieved underscores respiratory contaminations and demanding medical
S.A.; Approach for learning an the efficacy attention on a global scale. It has had a profound effect
Palanikku COVID-19 approach called accuracy of deep on people lives, society, and the global economy. Due of
mar D. Detection Using Self-Trans. of 98.7% learning the severe limitations of RT-PCR (switch record
CT Scans methodologi polymerase chain reaction)-based methods for COVID19
es, diagnosis, various healthcare-based imaging approaches
specifically have lately been used.
Self-Trans, in
enhancing
COVID-19
diagnosis
through the
utilization of
CT scans.
Wang J.; A Reconfigurable The study The model The study This work presents methods and architectures for scaling
Dang Hardware utilizes deep achieved showcases DL inference for pneumonia detection in chest X-ray
K.N.; Platform for learning an how images based on a reconfigurable self-contained
Abdallah Efficient Deep techniques for accuracy proposed hardware platform named AIRBiS1. The performance
A.B. Learning pneumonia of 95.2%. methods evaluation results show that the proposed approach
Inference in detection in and achieves 95.2% detection accuracy of pneumonia over
Pneumonia chest X-ray architecture the collected test data with the computer-aided diagnosis
Detection from images on the s effectively scenario.
Chest X-ray AIRBiS1 tackle
Images platform. latency,
security, and
power
consumption
challenges
in
healthcare-
focused AI
applications.
Mall P.K.; DL uses a large- false discovery
Singh scale labelled rate, false-
P.K. dataset to positive rate, f1
ensure a high score, negative
degree of predictive value,
accuracy. This and false-
technology is negative rate
increasingly 91.834%,
data-driven in 85.268%,
medicine and 97.008%,
biology imaging, 89.356%,
and labelled data 83.648%,
is more difficult 10.644%,
and expensive to 14.732%,
 retrieve. Various 93.016%,
studies are being 95.696%, and
conducted on 2.992% for
semi-supervised unseen dataset
deep learning respectively. This
models (SSDLM) research work
and self- leads to a more
supervised deep accurate and
learning. In order efficient SSDLM.
to increase the
quantity of
labelled data
necessary for
deep learning,
researchers are
increasingly
looking at SSDLM
and its
applications.
Mir M.H.; The lack of Based on the
Jamwal modern experimental
S.; Iqbal technology in results, our
U.; healthcare has proposed models
Mehbodn led to the death have recorded
iya A.; of thousands of good accuracies
Webber lives worldwide with highest of
J.; Khan due to COVID-19 97.767% by
U.H. since its Support Vector
outbreak. The Machine.
IoT along with According to the
other findings of
technologies like experiments, the
Machine proposed
Learning can conceptual
revolutionize the model will aid in
traditional the early
healthcare detection and
system. Instead prediction of
of reactive COVID-19
healthcare suspects, as well
systems, IoT as continuous
technology monitoring of the
combined with patient in order
machine learning to provide
and edge emergency care
computing can in case of
deliver proactive medical volatile
and preventive situation.
healthcare
services. In this
 study, a novel
healthcare edge-
assisted
framework has
been proposed
to detect and
prognosticate
the COVID-19
suspects in the
initial phases to
stop the
transmission of
coronavirus
infection.
Ahmed Lung cancer is these evaluation
S.; Raza the most performance
B.; common and measures. The
Hussain second leading results revealed
L.; cause of cancer that proposed
Aldweesh with lowest approach is more
A.; Omar survival rate due robust than
A.; Khan to lack of traditional ML
M.S.; efficient algorithms.
Eldin diagnostic tools. Based on these
E.T.; Currently, results, the
Nadim researchers are proposed
M.A. devising AI methodology can
based tools to be very helpful in
improve the the early
diagnostic detection and
capabilities. The treatment of
machine learning lung cancer for
(ML) requires better diagnosis
hand-crafted system.
features to train
the algorithms.
To extract most
relevant features
is still a
challenging task
in the field
image
processing. We
first extracted
the texture gray
level co-
occurrence
matrix features.
We fed these
features to
 traditional
Lal J.D.; Security is the Followed by this,
Ayoub S.; major concern the GSO
Hakim while seeing the approach is
P.D.; dramatic utilized for the
Prabagar expansion in its optimum
S.; applications and hyperparameter
Dwivedi size. It is adjustments of
V.K.; challenging to the DAE model,
Tiwari M. independently which leads to
put security enhanced attack
mechanism in all detection
the IoT devices efficacy. The
and upgrade it experimental
according to evaluation of the
newer threats. GSODL-ADC
Furthermore, algorithm is
machine learning tested against
(ML) techniques benchmark
could better dataset. The
apply the obtained
massive quantity experimental
of data produced values signify
by IoT devices. the betterment
Thus, several DL of the GSODL-
based algorithms ADC technique
were introduced for attack
for detecting recognition
attacks in IoT. purposes.
Ajwad Secure IoT have Also, an AI
A.A.; evolved into a technique is
Ahmed requirement for applied for
A.A.; electronic COVID-19 testing
Kamal healthcare such as adopted
M.; Jaleel systems. In most a convolutional
R.A.; cases, health neural network.
Mahmoo images contain Evaluation
d M.B. sensitive showed that the
information proposed
about patients framework
that must be outperformed
protected. alternative
Traditional security and
encryption transfer learning
cannot be methodologies in
directly applied secure
to image data prediction.
due to Therefore, it can
restrictions in be used to
digital data reliably transmit
 attributes. CT-images for
Additionally, COVID-19
patients may patients while
lose the meeting strict
confidentiality of security and
their data when prediction
private images benchmarks.
are transmitted
via a network.
Thus, multimedia
AI and image
processing are
applied to build
improved secure
IoTs.
Amellal This paper information
A.; presents a issue, and the
Amellal solution to a relevance of the
I.; challenge faced methods used
Seghioue in the supply has been verified
r H.; Ech- chain via precise
Charrat management of performance
M.R. a spare parts indicators, such
distributor with a as the RMSE
dispersed global (Root of the
supply network mean of the
and local square of the
distribution errors) testifying
network in to the accuracy
Morocco. The of the results
problem is a lack and also through
of accurate lead comparison with
time information, other models.
leading to
difficulties in
meeting
customer
demand. The
proposed
solution is a
framework using
an LSTM model
for lead time
Awotund IoT and ML have The findings
e J.B.; recently changed from the study
Misra S.; the paradigm in show that the
Ajagbe the area of XGBoost
S.A.; Ayo applications in performs reliably
F.E.; health-care better when
Gurjar R. systems with compared with
 significant other existing
results. The ML- ML-based models
based models with an accuracy
enabled with IoT- of 95.7%, recall
based systems of 95.9%,
have greatly precision of
expedited the 96.0%, F1-score
medical of 94.8%, and
diagnosis, real- Receiver
time monitoring, Operating
and treatment in Characteristic
the health-care Curve (ROC) of
system, thus 94.3%. The
reducing the results proposed
burden of health- a framework to
care processes show that the
and the cost of system can be
taking care of widely used to
patients by diagnose various
caregivers and diseases and can
physicians. be used in
remote areas for
better
monitoring of
patients.
Mohapatr Automated Face CNN, SSD, Faster The The study After the global spread of the epidemic, especially in less
a S.K.; Ali Mask Detection R-CNN system's demonstrate developed countries, the World Health Organization
F.A.; Using Python, ability to s the (WHO) classified the Novel Coronavirus (COVID-19) as a
Sarangi Deep Learning, detect effectivenes severe threat to public health. In response, people are
P.K.; and Computer face mask s of the
advised to limit social contact, frequently wash their
Sahu P.; Vision for COVID- usage is developed
hands, and wear masks due to limited antiviral
Mohanty 19 Safety assessed mask
treatments and healthcare resources.
J. Monitoring by detection
comparing approach
it with using deep
other learning and
methods computer
using vision
metrics techniques.
like
accuracy,
precision,
recall, and
F1-score.
Nguyen- Enhancing CNN-biLSTM, C- The The study Early detection of the disease is an important factor to
Trong K.; Accuracy in YAMNet, accuracy demonstrate reduce its effect and quickly control the spread of
Nguyen- COVID-19 RNNoise. score of s the pandemic. Nevertheless, many countries still rely on
Hoang K. Diagnosis: the effectivenes methods that are either expensive and time-consuming.
Integrating proposed s of a multi- Recently, deep learning methods have been proposed as
Cough Sounds method modal a potential solution for COVID-19 analysis. However,
 and Symptom on the approach for previous works usually focus on a single symptom, which
Reports with testing set COVID-19 can omit critical information for disease diagnosis.
Deep Learning is 96.9% detection Therefore, in this study, we propose a multi-modal
Techniques using deep method to detect COVID-19 using cough sounds and self-
learning reported symptoms.
techniques,
including
cough
sounds and
self-reported
symptoms
as input
features.
Jothiarun Cardiovascular By incorporating
a N.; diseases hyperparameters
Leema accurate such as learning
A.A. detection and rate, optimizer,
classification at and dropout, the
an early stage, classification
aiming to accuracy can be
mitigate the enhanced. These
severity of the hyperparameters
condition. The play a crucial
abrupt role in fine-
obstruction of tuning the
blood vessels is classification
identified as the model and
leading cause of improving its
fatality in a short overall
timeframe. performance.
Consequently, it The accuracy of
is imperative to our model
promptly reached the
diagnose and highest level at
initiate learning rate of
treatments to 0.0001, dropout
effectively is 0.3 and the
address the Adam optimiser.
disease. LW-DN161 model
Electrocardiogra attained 95.74%
m (ECG) images accuracy,
are used. The 95.77%
acquired dataset precision,
is unbalanced, 95.75% recall,
therefore while and 95.77% F1-
training the Score.
model it will lead
to an overfitting
issue.
Sandhu Using the It can completely
J.K.; Kaur Internet of change how
A.; Things (IoT) and hospitals work to
Kaushal Machine find and predict
C. Learning (ML), it cancers with
is possible to find higher care and
breast cancer customer
quickly and satisfaction. The
accurately, analysis found
which is that the SVM
important for method had the
getting better highest accuracy
and getting (98.1%)
treatment right compared to the
away. In recent other methods.
years, the IoT The purpose of
has changed in a this study is to
way that makes explore the
it possible to use application of
artificial imaging and
intelligence (AI) machine learning
and ML to look at in conjunction
both real-Time with internet-of-
and historical Things systems
data. Many in the detection
imaging of breast cancer
methods, such at an early
as stage.
mammography,
thermography,
CT,
Singh P.; The most extreme
Gupta S.; commonly gradient-
Gupta V. occurring cancer boosting, light
among women, gradient-
breast cancer, boosting
causes lakhs of machine and
deaths annually, categorical
which can be boosting on
prevented by features
early detection obtained from
and treatment. Inception-
Detection can be ResNet-v2 CNN
done by using model applied on
machine learning the benchmark
models on BreakHis dataset
histopathological to optimize
images which Precision, Recall,
are affordable, Accuracy, and
reliable, and AUC
accurate. simultaneously.
 Previous studies The novel
in this regard NSGA2-IRv2-CXL
have focused on model proposed
transfer learning in this study
methods achieves
combining maximum
feature selection Accuracy of
using 94.40%, AUC of
Convolutional 98.16, Precision
Neural Networks of 95.77%, and
(CNNs) and an Recall of 99.29%
ensemble of for 100 ×
gradient- magnification.
boosting The study also
algorithms. establishes
trade-offs
between
performance
metrics thereby
opening avenues
for further
research in
multi-objective
approaches to
BCD which can
provide a larger
view of the
strengths and
weaknesses of
the classification
model.
Mahfuri Nowadays, the
M.; integration of
Alnawais blockchain and
eh A.E.; edge computing
Asad U.; has become
Yousif M.; important in
Aslam S.; computer
Chabani science due to
Z.; Fiaz the significant
M. advantages it
can bring to a
range of real-
world
applications. By
combining the
strengths of both
technologies, it
is Combining the
strengths of both
 technologies
makes it possible
to create a
highly secure
and reliable
network with
reduced latency
and efficient
transaction
processing.
Arji G.; The worldwide However, the
Ahmadi spread of the practical
H.; COVID-19 application of
Avazpoor disease has had these emerging
P.; a catastrophic tools for
Hemmat effect on managing
M. healthcare disturbance and
supply chains. warranting
The current resilience in the
manuscript supply chain has
systematically been examined
analyzes existing only rarely. This
studies article provides
mitigating directions for
strategies for additional
disruption research, which
management in can guide
the healthcare researchers to
supply chain develop and
during COVID-19. conduct
Using a impressive
systematic studies related to
approach, we the healthcare
recognized 35 supply chain for
related papers. different
Artificial disasters.
intelligence (AI),
block chain, big
data analytics,
and simulation
are
Hussen Real-time deep The six-layer
N.; learning faces FSBDL
Elghamra the challenge of framework
wy S.M.; balancing includes data
Salem accuracy and pre-processing,
M.; El- time, especially feature selection,
Desouky in cybersecurity hyper-
A.I. where intrusion parallelism, a
detection is customized CNN,
 crucial. a GUI layer for
Traditional deep interpretation,
learning and a detection-
techniques have evaluation layer.
been insufficient The optimized
in identifying CNN was
network designed to
anomalies and detect intrusions
intrusions. To in real-time
address this, a without
Fully Streaming compromising
Big Data accuracy or
Framework reliability. The
based on proposed
optimized Deep algorithms were
Learning for evaluated using
cybersecurity various
(FSBDL) was performance
proposed. The metrics, showing
framework that the
leverages two accuracy of the
parallel framework
optimization surpasses
algorithms, 99.9%, indicating
Adam and its superiority
RMSprop, labeled over other
Hyper-parallel intrusion
optimization detection
(HPO) techniques models. This
to enhance novel intrusion
efficiency and detection model
stability. offers promising
prospects for
cybersecurity,
and its
effectiveness
and accuracy
have been
demonstrated
through
experimentation.
Fiaz M.; Improving CNN The The study These models were developed to enhance confidence in
Al- Precision in proposed demonstrate the quality of 4D CBCT images. Accessing a large dataset
Khasawn Radiation CNN s how the was crucial for this task, made feasible by advancements
eh M.A.; Therapy models proposed in 4D CBCT research. Each phase-resolved image
Irtaza G. Targeting outperfor CNN models generated by the N-Net network, currently in
Tumors: m existing effectively development, is expected to surpass its predecessor
Enhancing 4D methods improve 4D reconstructed by U-Net using full projection data. This
CBCT Image in CBCT image
comparison demonstrates the effectiveness of the N-Net
Quality through reducing quality
 Noise Reduction streaking across technique in achieving maximum accuracy in
and Streaking artifacts diverse reconstructed images.
Aberration and noise patient
Reduction in 4D datasets and
CBCT imaging
images technologies
while , addressing
preserving associated
distinguis challenges
hing with broad
features. applicability

Kaur G.; The growth of The technical

Adhikari the fish is specifics of DL
N.; influenced by a approaches used
Krishnapr variety of in smart fish
iya S.; scientific factors. farming which
Wawale So, profit can be includes data
S.G.; easily achieved and algorithms
Malik by using some as well as
R.Q.; clever performance was
Zamani techniques, for also examined.
A.S.; example, In a nutshell, our
Perez- maintaining the goal is to provide
Falcon J.; correct pH level academics and
Osei- along with the practitioners with
Owusu J. dissolved oxygen a better
(DO) level and understanding of
temperature, as the current state
well as turbidity of the art in DL
for good growth implementation
of fish. Fully in aquaculture,
grown fish are which will help
generally sold at them deploy
a good price smart fish
because price of farming
fish in the applications as
market is well their
governed by benefits.
weight as well as
size of nurtured
fish.
Narayan Real-time In Enhance-Net
V.; Mall medical image models, the
P.K.; classification is a current study
Alkhayya complex problem combines image
t A.; in the world. enhancement
Abhishek Using IoT and green
K.; technology in channel with
Kumar medical Champion-Net. In
S.; applications the final step,
Pandey P. assures that the radiologists and
healthcare orthopaedic
sectors improve surgeons use the
the quality of trained model for
treatment while real-time medical
lowering costs image prediction.
via automation The study effort
and resource uses the
optimization. musculoskeletal
Deep learning is radiograph-bone
critical in classification
categorizing (MURA-BC)
medical images, dataset.
which is Classification
accomplished by accuracy of
artificial Enhance-Net was
intelligence. determined for
Deep learning the train and test
algorithms allow datasets. These
radiologists and models obtained
orthopaedic 98.02 percent,
surgeons to 94.79 percent,
make their life and 94.61
easier by percent
providing them accuracy,
with quicker and respectively. The
more accurate 96.74% accuracy
findings in real was achieved
time. Despite during real-time
this, the classic testing with the
deep learning unseen dataset.
technique has hit
its performance
limits. For these
reasons, in this
research, we
examine
alternative
enhancement
strategies to
raise the
performance of
deep neural
networks to
provide an
optimal solution
known as
Enhance-Net.
Kshirsag Examination and E-business
ar P.R.; evaluation of Associations may
Reddy sentiment The track the uses
D.H.; most popular and passions
Dhingra area for associated with
M.; dissecting and their Particulars
Dhabliya extracting via this internal
D.; Gupta knowledge from assessment and
A. communication adopt better
data from many marketing
sources, such as mechanism to
Facebook, give a tailored
Instagram, shopping
Twitter, Amazon, experience for
and so on, is their customers,
known as so improving
mining. It plays a their hierarchical
fundamental role advantage. By
in enabling the using the
organizations to Programming
work interface key,
productively on Twitter data for
advancing the this article was
business system collected from
and gaining the Twitter. Similarly,
most we really want to
comprehensive finish setting up
understanding of the systems.
the buyer's Additionally, the
feedback on system
their product. supported NLP
approaches, thus
we wish to
implement the
calculation as a
crucial backslide.
The exploratory
problems exhibit
perfection.
Alamelu Recently image An Improved
V.; processing YOLO-v4 with
Thilagam techniques CSPDarkNet53 is
ani S. require a clear used for feature
focused solution extraction and
to predict classifying the
diseases. The echo-cardiogram
primary goal of pictures. Both
the proposed categorization
method is to use approaches were
health used, and the
information and results were
 medical pictures integrated and
for classifying confirmed in the
the data and ability to forecast
forecasting heart disease.
cardiac disease. The LBO-YOLO-v4
It consists of two process detected
phases for regular sensor
categorizing the data with
data and 97.25% accuracy
prediction. If the and irregular
previous phase’s sensor data with
results are 98.87%
practical heart accuracy. The
problems, then proposed
there is no need improved YOLO-
for phase 2 to v4 with the
predict. CSPDarkNet53
method gives
better
classification
among echo-
cardiogram
pictures.
Bharti V.; ML, DL and AI and optimal
Biswas enable architecture are
B.; automating the just a few of the
Shukla computer-aided issues that can
K.K. system and be efficiently
drastically addressed with
reducing the swarm
human workload intelligence,
where correct whereas in the
prediction with case of swarm
accurate intelligence, slow
precision is convergence,
needed. On the local optima
other side, as stagnation, and
technology extensive
advances, a vast computation cost
amount of data can be
is generated, addressed with
raising the the ML and DL
problem community.
complexity and Therefore, a
computational robust and self-
challenges of efficient model
real-world can be
applications. developed by
Furthermore, integrating these
the ML and DL, concepts to solve
 the majority of the complex
real-world problem
applications associated with
have complex real-world
optimization applications. This
problems within hybrid approach
themselves that benefits the
must be majority of
adequately research
addressed for domains. Thus,
better and more this chapter will
accurate primarily present
analysis. the ideas,
Nonetheless, we challenges, and
believe that recent trends of
swarm an integrative
intelligence- approach of
based swarm
approaches to intelligence with
deep learning deep learning,
have which is
traditionally currently in high
been demand for
understudied addressing
and may industrial
ultimately problems.
deliver similar
advances in AI
capabilities -
either building
on those
provided by deep
learning or
offering whole
new ones.
Dutta D.; The present Transfer learning
Pal S.K. study focuses on aims to utilize
the prediction the information
and assessment gained from one
of the impact of problem to
lockdown improve the
because of predictive
coronavirus performance of a
pandemic on the learning model
air quality during for a different
three different but related
phases, viz., problem. Our
normal periods work helps to
(1 January 2018– demonstrate
23 March 2020), how TL is useful
 complete when there is a
lockdown (24 scarcity of data
March 2020–31 during the
May 2020), and COVID-19
partial lockdown pandemic
(1 June 2020–30 regarding the
September drastic change in
2020). We concentration of
identify the most pollutants. The
important air results reveal the
pollutants best prediction
influencing the performance of
air quality of TLS-BDLSTM with
Kolkata during a lead time of
three different 24 h as
periods using compared to
Random Forest, some well-known
a tree-based ML traditional ML
algorithm. It is and statistical
found that the models and the
ambient air pre-trained
quality of Kolkata stacked-BDLSTM.
is mainly The prediction is
affected with the then validated
aid of particulate using the real-
matter or PM time data
(PM10 and obtained during
PM2.5). the complete
However, the lockdown due to
effect of the COVID second
lockdown is most wave (16 May–15
prominent on June 2021) with
PM2.5 which different time
spreads in the air steps, e.g., 24 h,
of Kolkata due to 48 h, 72 h, and
diesel-driven 96–120 h.
vehicles,
domestic and
commercial
combustion
activities, road
dust,
Afriyie Y.; Medical anomaly
Weyori identification
B.A.; using machine
Opoku learning is a
A.A. significant
subject that has
received a lot of
attention.
 Artificial neural
networks’
successor, deep
learning, is a
well-developed
technology with
strong
computational
capabilities. Its
popularity has
increased in
recent years due
to the availability
of rapid data
storage and
hardware
parallelism.
Numerous,
sizeable medical
imaging datasets
have recently
been made
available to the
public, which has
sparked interest
in the field and
increased the
number of
research studies
and publications.
The main goal of
this study is to
give a complete
theoretical
examination of
prominent DL
algorithms for
detecting
medical
anomalies.
Swain Exploring ANN The The study COVID-19 disease is rigorously validated using statistical
B.K.; Gammatone accuracy concludes measures, F1 score, confusion matrix, specificity, and
Khan Frequency score for that GFCC, sensitivity parameters. Besides, it is found that the
M.Z.; Cepstral the when used proposed algorithm using GFCC performs well in terms of
Chowdha Coefficients proposed with the detecting the COVID-19 disease from the noisy
ry C.L.; (GFCC) for algorithm DHO-ANN crowdsourced cough dataset, COUGHVID. Moreover, the
Alsaeedi COVID-19 using algorithm, proposed algorithm and undertaken feature parameters
A. Detection Using GFCC in yields better have improved the detection of COVID-19 by 5%
DHO-ANN: detecting results for compared to the existing methods.
Analysis of Noisy COVID-19 COVID-19
 Crowdsourced from the detection
Cough Datasets noisy than the
crowdsour commonly
ced cough used Mel-
dataset frequency
COUGHVI cepstral
D is 5% coefficient in
higher noisy
compared crowdsource
to existing d speech
methods. corpora.
Guadiana Development of SMOTE, KNN. The The study Additionally, the benefits of the data imputation
-Alvarez, a COVID-19 accuracy concludes algorithm for unavailable biomarker data are highlighted.
J. L., Mortality Risk score for that the DL- The proposed scheme is deemed suitable for critically ill
Hussain, Calculator Using the based COVID-19 patients to assess the risk of mortality and
F., Deep Learning: proposed COVID-19 prognosis effectively.
Morales- Enhancing method mortality
Menende Patient on the risk
z, R., Management dataset calculator
Rojas- and Resource comprisin significantly
Flores, Allocation g 1,503 outperforms
E., seriously existing
García- ill COVID- methods in
Zendejas 19 predicting
, A., patients is mortality
Escobar, 95%. risk for
C. A., COVID-19
Ramírez- patients.
Mendoza
, R. A., &
Wang, J.
Kamnard Recent The results show
siri T.; advancement in that YOLO v5
Charoenk DL-based works best for
wan P.; Convolutional the ParcelBar in
Malang Neural Networks terms of speed
C.; (D-CNNs) has led and accuracy.
Wudhikar research to The situation is
n R. improve the different for the
efficiency and InventBar since
performance of Faster R-CNN
barcode could allow the
recognition in model to learn
Supply Chain faster with a
Management small drop in
(SCM). D-CNNs accuracy. It is
required real- proven that the
world images proposed
embedded with datasets can be
ground truth practically
 data, which is utilized for the
often not readily mainstream D-
available in the CNN frameworks.
case of SCM Both are
barcode available for
recognition. This developing
study introduced barcode
two invented recognition
barcode models and
datasets: positively affect
InventBar comparative
studies.
Chou J.- To develop an AI To enhance the
S.; model to generalization
Cheng T.- forecast PMFC performance of
C.; Liu C.- power CNN, a newly
Y.; Guan generation developed bio-
C.-Y.; Yu accurately, inspired
C.-P. relevant results optimization
obtained using algorithm,
shallow and DL jellyfish search
techniques are (JS), was
compared for the incorporated into
first time. Once this model to
deep learning determine the
techniques had optimal
been identified hyperparameters
as superior for , yielding the
this purpose, hybrid JSCNN
they were used model. This
with a bio- investigation
inspired revealed that the
optimization JS optimization
algorithm to algorithm can
dynamically find better
setting the values of
model hyperparameters
hyperparameters of the CNN and
. The developed stabilize model
model can also accuracy.
be applied to Notably, once
estimate the the optimal
power hyperparameters
generation have been
capacity of PMFC obtained using
devices in the JS, the
future. The computation
model was time of the
trained using hybrid JSCNN
data collected model is shorter
 from sensors in a than that of the
site experiment generic CNN
that was carried model,
out using PMFCs supporting the
embedded with need to
Chinese determine the
pennisetumin, appropriate
narrowleaf hyperparameter
cattail, dwarf values in deep
rotala, and no learning. This
plant as a control study is the first
group. The to use its
original data of particular setup
device and to offer its
parameters, particular
environmental precautions; it
parameters, and therefore contrib
the utes to the body
of domain
knowledge and
practicality of
sustainable
PMFCs.
ElDahsha Numerous 98.93, 99.63,
n K.A.; network 99.19, 99.78,
AlHabshy cyberattacks and 0.01, with
A.A.; have been 0.51 for average
Hameed launched due to accuracy, DR,
B.I. inherent and FAR in the
weaknesses. UNSW-NB15 and
Network CICIDS2017
intrusion datasets,
detection is a respectively.
crucial
foundation of the
cybersecurity
field. Intrusion
detection
systems (IDSs)
are a type of ML
software
proposed for
making decisions
without explicit
programming
and with little
human
intervention.
Although ML-
based IDS
 advancements
have surpassed
earlier methods,
they still struggle
to identify attack
types with high
detection rates
(DR) and low
false alarm rates
(FAR). This paper
proposes a meta-
heuristic
optimization
algorithm-based
hierarchical IDS
to identify
several types of
attack and to
secure the
computing
environment.
The proposed
approach
comprises three
stages: The
Gasmi K. Automatic
classification of
texts is a well-
known topic in
natural language
processing (NLP),
and it involves
categorizing
unstructured
texts into
specific groups.
Classification
algorithms for a
given problem
are searched for,
as well as the
appropriate
features to use
as input. These
models have
been used for
analysis of data
with a large
number of
sparse
 dimensions. In
medical field,
categorization of
medical texts is
a difficult task
because these
types of texts
contain several
terminologies
that define
medical ideas
and words.
Medical data also
lack proper
sentence
structure and do
not adhere to the
standard rules of
natural language
grammar. The
main problem in
this classification
task is to choose
an appropriate
representation
as input. To solve
the
representation
problem, we
propose a new
extension model
for
Das B. Although some Support vector
people do not machine (SVM)
have any chronic and k nearest
disease or are neighborhood (k-
not in the risky NN) are chosen
age group for as classifiers.
Covid-19, they The method
are more achieved the
vulnerable to the highest
coronavirus. As classification
the reason for accuracy rate of
this situation, 98.84% with an
some experts SVM classifier to
focus on the detect Covid-19
immune system from DNA
of the person, signals. The
while others proposed
think that the method is ready
 genetic history of to be tested with
patients may a different
play a role. It is database in the
critical to detect diagnosis of
corona from DNA Covid-19 using
signals as early RNA or other
as possible to signals.
determine the
relationship
between Covid-
19 and genes.
Thus, the effect
on the severe
course of the
disease of
variations in the
genes associated
with the corona
disease will be
revealed. In this
study, a novel
intelligent
computer
approach I
Iqbal S.; DL is an obvious classification
Qureshi method for the results were
A.N.; detection of obtained using
Ullah A.; disease, large, publicly
Li J.; analyzing available clinical
Mahmoo medical images datasets such as
d T. and many BreakHis, BraTS,
researchers have NIH-Xray and
looked into it. COVID-19 X-ray.
However, the Medical
performance of practitioners and
deep learning clinicians can
algorithms is utilize the CNN
frequently model to
influenced by corroborate their
hyperparameter first malignant
selection, the and benign
question of classification
which assessment. The
combination of recommended
hyperparameters Adaptive high F1
are best score and
emerges. To precision, as well
address this as its excellent
challenge, we generalization
proposed a novel and accuracy,
 algorithm for imply that it
Adaptive might be used to
Hyperparameter build a
Tuning (AHT) that pathologist’s aid
automates the tool.
selection of
optimal
hyperparameters
for Convolutional
Neural Network
(CNN) training.
All of the optimal
hyperparameters
for the CNN
models were
instantaneously
selected and
Jamei M.; Particulate ML methods (i.e.,
Ali M.; matter (PM) or LightGBM and
Malik A.; particle pollution ridge kernel
Karbasi include the tiny regression (KRR))
M.; particles of dust and two
Sharma and fly ash traditional ML
E.; particles are methods (i.e.,
Yaseen expelled from Adaptive neuro-
Z.M. coal-burning fuzzy inference
power plants. system (ANFIS)
Coal combustion and multilayer
is an extremely perceptron
prevalent source neural network
of air pollution, (MLP)) were
and resulting PM hybridized with
has substantial the multi-level
impacts on data filtering
human health, strategy by
especially in examining the
industrial zones. optimal input
This paper aims combinations.
to design a novel Several
hybrid deep statistical
learning metrics,
framework based graphical tools,
on long short- and diagnostic
term memory analyses
(LSTM) evaluated the
integrated with a hybrid models.
two-stage data
filtering
technique to
accurately
 predict the air
quality indices
(i.e., PM2.5 and
PM10) in a
chosen study
region ‘Miles
Airport,
Queensland,’
that meets the
needs of the coal
seam gas
industry in
Australia. The
data used to
construct the
novel hybrid two-
stage data
filtering
technique
Wei J.; Industrial
Zhu Q.; Internet of
Li Q.; Nie Things (IIoT)
L.; Shen facilitate private
Z.; Choo data collecting
K.-K.R.; via (a broad
Yu K. range of)
sensors, and the
analysis of such
data can inform
decision making
at different
levels. Federated
learning (FL) can
be used to
analyze the
collected data, in
privacy-
preserving
manner by
transmitting
model updates
instead of
private data in
IIoT networks.
The FL
framework is,
however,
vulnerable
because model
updates are
Hussein Chest and lung the proposed
F.; diseases are method
Mughaid among the most implemented the
A.; serious chronic support vector
AlZu’bi diseases in the machine to
S.; El- world, and they classify the
Salhi occur as a result images with and
S.M.; of factors such without using
Abuhaija as smoking, air CLAHE equalizer.
B.; pollution, or The obtained
Abualiga bacterial results were
h L.; infection, which compared with
Gandomi would expose the CNN
A.H. the respiratory networks. Later,
system and two different
chest to serious experiments
disorders. Chest were
diseases lead to implemented
a natural with hybrid
weakness in the architecture of
respiratory deep CNN
system, which networks and
requires the CLAHE as a
patient to take preprocessing for
care and image
attention to enhancement.
alleviate this The
problem. experimental
Countries are results indicate
interested in that the
encouraging suggested hybrid
medical research architecture
and monitoring outperforms
the spread of traditional
communicable methods by
diseases. roughly 20% in
terms of
accuracy.
Vinoth This paper The
R.; proposes an experimentation
Ananth effective is done by using
J.P. classification the Switzerland
method named dataset,
Rider Chicken Cleveland
Optimization dataset,
Algorithm-based Hungarian
Recurrent Neural dataset and Skin
Network (RCOA- disease dataset,
based RNN) to in which the
perform big data proposed RCOA-
 classification in based RNN
spark attained better
architecture. performance
Initially, the based on the
input data are quantitative
collected from properties, such
the network by as sensitivity,
the master node accuracy and
and then specificity with
forwarded to the the values of
slave node. 9.3E+01%,
These nodes are 9.4E+01% and
responsible for 9.3E+01% using
storing the data Hungarian
and performing dataset. The
existing learning
methods failed
to address the
complex
classification
problems at a
reasonable time,
which is
overcome by the
proposed
method.
Shi B.; Ye COVID-19 is an classification The
H.; extreme acute performance of statistical
Heidari respiratory the EBSO-SVM is results
A.A.; syndrome. Early very promising, demonstra
Zheng L.; diagnosis and reaching te that the
Hu Z.; accurate 91.9195% proposed
Chen H.; assessment of accuracy, EBSO-SVM
Turabieh COVID-19 are not 90.529% shows
H.; available, Matthews predictive
Mafarja resulting in correlation properties
M.; Wu P. ineffective coefficient, for all
therapeutic 90.9912% metrics
therapy. This Sensitivity and and higher
study designs an 88.5705% stability,
effective Specificity on which can
intelligence COVID-19. be treated
framework to Compared with as a
early recognition other existing computer-
and state-of-the-art aided
discrimination of methods, the technique
COVID-19 EBSO-SVM in this for
severity from the paper still shows analysis of
perspective of obvious COVID-19
coagulation advantages in severity
 indexes. The multiple metrics. from the
framework is perspectiv
proposed by e of
integrating an coagulatio
enhanced new n.
stochastic
optimizer, a
brain storm
optimizing
algorithm
(EBSO), with an
evolutionary
machine
Ascencio In this work, the Friedman–
-Cabral performance of Nemenyi test.
A.; five deep The results show
Reyes- learning that less
Aldasoro architectures in complex
C.C. classifying architectures
COVID-19 in a such as ResNet-
multi-class set- 50, ResNet-50r
up is evaluated. and DenseNet-
The classifiers 121 were able to
were built on achieve better
pretrained generalization
ResNet-50, with rankings of
ResNet-50r (with 1.53, 1.71 and
kernel size 3.05 for the
(Formula Matthew
presented.) in Correlation
the first Coefficient,
convolutional respectively,
layer), while MobileNet-
DenseNet-121, v3 and CaiT
MobileNet-v3 obtained
and the state-of- rankings of 3.72
the-art CaiT-24- and 5.0,
XXS-224 (CaiT) respectively.
transformer. The
cross entropy
and weighted
cross entropy
were minimized
Bassiouni Utilizing Deep CNNs, Vgg16, The The study The study addresses the critical need for improved early
, M. M., Learning Vgg19, ResNet- accuracy showcases diagnosis of COVID-19. It involves data collection, image
Hegazy, Techniques for 101, and score for the viability filtering, augmentation, feature extraction, and
I., Rizk, Automated Xception, TCN. ECGConvn of deep classification of ECG images. Data from two publicly
N., El- COVID-19 et-SVM on learning with available datasets, including COVID ECG reports, undergo
Dahshan, Diagnosis from multi-class ECG data for preprocessing and class-balanced augmentation. A CNN-
E.-S. A., ECG Data tasks is COVID-19
& Salem, 99.74%, diagnosis. based approach extracts features, utilizing pre-trained
A. M. and for ECGConvnet models Vgg16, Vgg19, ResNet-101, and Xception. An
one , especially ensemble model, ECGConvnet, combines Xception and
binary- with SVM, temporary convolutional network (TCN). Results are then
class yields fed into softmax, RF, MLP, and SVM classifiers for
diagnosis, promising evaluation.
it is results,
99.8%. hinting at
the potential
for an
automated
COVID-19
diagnosis
system
using ECG
data.
Hu Y.; Xu Mask-face This was
Y.; detection has achieved by
Zhuang been a treating the
H.; Weng significant task incorrect-mask-
Z.; Lin Z. since the wearing scenario
outbreak of the as an anomaly,
COVID-19 leading to
pandemic in reasonable
early 2020. performance in
While various the absence of
reviews on mask- training data of
face detection the third class.
techniques up to
2021 are
available, little
has been
reviewed on the
distinction
between two-
class (i.e.,
wearing mask
and without
mask) and three-
class masking,
which includes
an additional
incorrect-mask-
wearing class.
Moreover, no
formal review
has been
conducted on
the techniques of
implementing
 mask detection
models in
hardware
systems or
mobile devices.
Pang B.; As an essential The second layer Simulation
Low K.H.; enabler to meet is developed as a results
Lv C. these mixed-integer demonstra
challenges, pre- nonlinear te that the
flight 4D routes programming proposed
optimization is (MINLP) model to adaptive
required to optimize the conflict
conduct conflict decision resolution
detection and variables of the framework
resolution strategies successful
(CD&R) and to selected by the ly
generate first layer. To optimizes
conflict-free solve the the
flight routes proposed double- strategies
before layer used for
departure. optimization each type
Existing studies problem, we of flight
on strategic introduce and conflict,
deconfliction improve a novel which
cover several meta-heuristic subseque
types of stochastic fractal ntly
strategies such search (SFS) optimizes
as scheduling or algorithm with the 4D
rerouting. two major routes
However, a improvements of with
single type of a penalty-guided significant
strategy used to fitness function reductions
solve different and an in total
types of conflicts exploitation- operation
may lead to an exploration al cost,
unsafe and balancing number of
inefficient way of scheme. flight
conflict conflicts,
resolution. This and flight
paper proposes delays.
an adaptive The
decision-making improved
framework to stochastic
optimize the fractal
resolution search
strategies used (ISFS)
for different algorithm
types of conflicts is also
with explainable proved
mechanisms. effective
 and
reliable in
solving
the
proposed
optimizati
on
problem in
different
traffic
density
scenarios.
Butt High blood sugar measured using
M.M.; in diabetic various metrics
Iskandar patients is the and results are
D.N.F.A.; main source of compared with
Abdelha Diabetic recent
mid S.E.; Retinopathy approaches for
Latif G.; (DR). This affects DR detection.
Alghazo the blood vessels The proposed
R. within the retina. method provides
Manual detection significant
of DR is a performance
difficult task improvement in
since it can DR detection for
affect the retina, fundus images.
causing For binary
structural classification, the
changes such as proposed
Microaneurysms modified method
(MAs), Exudates achieved the
(EXs), highest accuracy
Hemorrhages of 97.8% and
(HMs), and extra 89.29% for
blood vessel multiclass
growth. In this classification.
work, a hybrid
Zhao D.; The ant colony The results fully
Liu L.; Yu optimization proved that
F.; algorithm is a ADNOLACO has
Heidari classical swarm accelerated the
A.A.; intelligence convergence
Wang M.; algorithm, but it speed and
Chen H.; cannot be used improved the
Muhamm for continuous convergence
ad K. class accuracy. The
optimization ability to find a
problems. A balance between
continuous ant local and globally
colony optimal solutions
 optimization is improved.
algorithm Also, to show
(ACOR) is that ADNOLACO
proposed to has some
overcome this practical value in
difficulty. Still, real applications,
some problems it deals with four
exist, such as engineering
quickly falling problems. The
into local simulation
optimum, slow results also
convergence illustrate that
speed, and low ADNOLACO can
convergence improve the
accuracy. To accuracy of the
solve these computational
problems, this results.
paper proposes a Therefore, it can
modified version be demonstrated
of ACOR called that the
ADNOLACO. proposed
There is an ADNOLACO is a
opposition-based promising and
learning excellent
algorithm based
on the results.
Han Z.; Enhancing Face YOLOv4-tiny. The The study's In this study, we propose an efficient and lightweight
Huang Mask Detection accuracy detector face mask detection method based on YOLOv4-tiny. Our
H.; Fan Accuracy in score for surpasses system monitors mask-wearing status in complex
Q.; Li Y.; Complex the state-of-the- environments, focusing on improving detection accuracy
Li Y.; Environments proposed art models, while maintaining real-time performance. Key
Chen X. Using YOLOv4- detector, improving enhancements include optimizing the network structure,
tiny as real-time applying the K-means++ clustering algorithm, and
indicated mask designing improved residual and cross fusion modules to
by the detection effectively extract features from small or medium-sized
mean performance targets.
average and
precision enhancing
(mAP) flexibility
increase, across
is 4.56%. diverse
settings like
hospitals,
campuses,
and
communities
.
Ketu S.; IoT is one of the integration of
Mishra emerging various IoT-based
P.K. technologies, application areas
 which is widely that play a key
used across the role in building a
globe. As the smarter city
idea of a smart have also been
city was discussed. This
founded, IoT has extensive study
been contributes a
acknowledged as useful panorama
the key on various key
foundation in points and gives
smart city a critical
paradigms. direction for
Technology forthcoming
makes a person investigations.
smart, and to This study will
make the world also provide a
smart, we have reference point
to make the for practitioners
country smart. To and academics in
make the the near future.
country smart,
we have to make
cities smart and
to make smart
cities, we have
to be smart. In
short, to create a
smart
environment,
one must be
equipped and
familiar with the
current trends.
The integration
of various smart
devices and
systems
facilitates IoT for
a smart city. The
Ghimire Developing a CNN, SVM. The CNN- The CNN- Our hybrid model is rigorously evaluated to analyze its
S.; Hybrid Prediction REGST REGST performance over a yearlong, and all four season data.
Nguyen- Model, CNN- model model is a We also compare the proposed CNN-REGST model with
Huy T.; REGST, for Daily demonstra valuable tool several deep learning (i.e., CNN, Long-term Short-term
Deo R.C.; Global Solar tes for Memory Network LSTM, Deep Neural Network DNN) and
Casillas- Radiation superior predicting conventional ML approaches (Extreme Learning Machine
Pérez D.; Prediction Using prediction Global Solar ELM, Stacked Regression REGST, Random Forest
Salcedo- Climate Models performan Radiation Regression RFR, Gradient Boosting Machine GBM,
Sanz S. and ce for (GSR), Multivariate Adaptive Regression Splines MARS) using the
(2022) Observations Global suitable for same test datasets.
Solar integration
 Radiation into modern
(GSR) solar energy
monitoring
technologies
to enhance
utilization of
sustainable
energy in
hybrid
energy grid
systems.
Turay T.; Layer-based overview of
Vladimiro details of CNNs recent ADSs
va T. along with related works
parameter and and current
floating-point industry
operation activities, aiming
number to bridge
calculations are academic
outlined. Using research and
an evolutionary industry practice
approach, the on CNNs and
majority of the ADSs. Design
outstanding approaches
image targeted at
classification solving problems
CNNs, published of automakers in
in the open achieving real-
literature, is time
introduced with a implementations
focus on their are also
accuracy proposed based
performance, on a discussion
parameter of design
number, model constraints,
size, and human vs.
inference speed, machine
highlighting the evaluations and
progressive trade-off analysis
developments in of accuracy vs.
convolutional size. Current
operations. technologies,
Results of a promising
novel directions, and
investigation of expectations
the convolution from the
types and literature on
operations ADSs are
commonly used introduced
in CNNs are including a
 presented, comprehensive
including a trade-off analysis
timing analysis from a human-
aimed at machine
assessing their perspective.
Liu J. n recent years, same time, the
due to the usability test
development of method is used
computer to compare and
technology and evaluate two
information common
technology, web computer-aided
technology has translation
changed the software, Trados
mode of and Wordfast. By
translation at an observing,
alarming rate. recording, and
The rapid analyzing user
development of behavior and
information related data, the
technology and five attributes of
globalization has usability can be
increased the learned and
demand for memorable. The
translation, experiments
especially show that the
technical effect of this
translation, and study on
the use of computer-aided
computer- software with the
assisted help of deep
translation learning
software can knowledge can
greatly improve produce good
the quality and results, and the
efficiency of robustness and
translation work. scalability of the
The purpose of software have
this article is that been enhanced,
under the increasing the
premise of competitiveness
continuous of the software
advancement in itself in
computer translation
technology, software.
computer-
assisted
translation can
effectively
improve the
 translation
efficiency of
Troisi O.; The spread of on the complex
Fenza G.; Covid-19 technology
Grimaldi profoundly acceptance
M.; Loia changed citizens' process and help
F. daily lives due to managers
the introduction identify the
of new modes of potential
work and access negative effects
to services based of smart
on smart technologies. In
technologies. this way, the
Although the study enhances
relevance of new scholars' and
technologies as practitioners'
strategic levers understanding of
for crisis the strategies for
resolution has enabling the use
been widely of technology
debated before within smart
the pandemic, cities to manage
especially in the the
smart cities' transformations
context, how introduced by
individuals have the health
agreed to include emergency and
the technological guide citizens’
changes dictated behaviour.
by the pandemic
in their daily
interactions
remains an open
question.
Rostami Developing The proposed The The study Several AI-based models have been developed for
M.; Explainable AI method utilizes developed demonstrate COVID-19 disease diagnosis. In spite of the promise of
Oussalah for COVID-19 an explainable model s the artificial intelligence, there are very few models which
M. Diagnosis from decision forest enhances potential of bridge the gap between traditional human-centered
Blood Test classifier for diagnosis using diagnosis and the potential future of machine-centered
Samples: COVID-19 accuracy explainable disease diagnosis. Under the concept of human-
Bridging Human- classification and artificial computer interaction design, this study proposes a new
Centered and based on patient reduces intelligence explainable artificial intelligence method that exploits
Machine- blood test data. execution methods for graph analysis for feature visualization and optimization
Centered Graph analysis is time COVID-19 for the purpose of COVID-19 diagnosis from blood test
Approaches employed for through diagnosis, samples. In this developed model, an explainable
feature novel offering a decision forest classifier is employed to COVID-19
visualization and feature transparent classification based on routinely available patient blood
optimization, selection. and test data.
enabling interpretable
clinicians to approach
 interpret the that
prediction model integrates
effectively. machine
learning with
human-
centered
diagnosis
practices.
Swain S.; ML has been underlying
Bhushan categorized as a systems for a
B.; branch of AI variety of
Dhiman under the wellness
G.; Computer applications, is
Viriyasita Science domain delineated by
vat W. wherein challenges and
programmable elevated by
machines imitate myriads of
human learning security. This
behavior with work tries to
the help of recognize a
statistical variety of
methods and vulnerabilities
data. The occurring in
Healthcare medical
industry is one of procurement,
the largest and admitting the
busiest sectors in concerns over its
the world, predictive
functioning with performance
an extensive from a privacy
amount of point of view.
manual Finally providing
moderation at possible risk
every stage. delimiting facts
Most of the and directions
clinical for active
documents challenges in the
concerning future.
patient care are
hand-written by
experts,
selective reports
are machine-
Kumar In recent years, learning analysis
A.; Khan with the method is used
S.F.; popularization of to automatically
Sodhi the Internet and give a multi-
R.S.; the development dimensional
Khan I.R.; of technologies score of the
Kumar such as big data doctor's service
S.; analysis, quality and
Tamrakar people's demand provide users
A.K. for mobile with the doctor
medical services and hospital
has become recommendation
more and more services. Finally,
urgent, which is by constructing
manifested in test sets and
determining their designing
diseases based questionnaires, it
on symptoms is verified that
and selecting the accuracy
hospitals with rates of disease
better service diagnosis service
quality according and doctor-
to the illnesses hospital
and doctors. An recommendation
inquiry service are
recommendation 74.00% and
system is 90.91 %,
designed and respectively.
implemented
based on
knowledge
graphs and deep
learning
technology to
solve the above
problems. Based
on the open
medical data on
the Internet, a
'disease-
symptom'
knowledge map
is constructed to
Balakesa To analyze Utilizing a pre- Analysis Social media Since the advent of COVID-19, lockdowns have been
va Reddy sentiment on trained model for includes platforms imposed all over the world. Hence, people have took to
P.; COVID-19 sentiment sentiment like Twitter the internet to voice their opinions and frustrations
Ramasub through Twitter analysis on distributio provide regarding this issue. Social media has become more
bareddy data during Tweets related to n, word valuable relevant after lockdown has been imposed as people are
S.; lockdown. COVID-19, cloud insights into confined to their homes and have no other means to stay
Viswanat categorizing visualizati public connected with family and friends. Hence, twitter will be
h G.; sentiments as on, and sentiment very insightful information medium to collect information
Govinda positive, neutral, trends in during crises on public opinion about COVID-19.
K. or negative. public like the
sentiment COVID-19
towards pandemic.
COVID-19 strategies.
over the
 course of
the
lockdown.
Azman Simplified Face CNNs is The The system In addition, the value of mask and unmask faces are
A.Z.; Mask Detection implemented. accuracy offers an saved in the cloud for future use. Using this DL enables
Zulaikha Using score efficient and the system to be faster and more precise to detect the
Saiful TensorFlow, achieved accurate faces, and as a result, the accuracy of mask and unmask
Bahri I.; Keras, and CNNs: for both method for faces detection is higher than 90%.
Saon S.; Enhancing Safety masked detecting
Mahama Measures During and face masks,
d A.K.; COVID-19. unmasked supporting
Anuarud faces is safety
din Bin over 90%. precautions
Ahmadon during the
M.; COVID-19
Yamaguc pandemic.
hi S.
Nasiri H.; Development of DenseNet169. The The In this paper, a pretrained network, DenseNet169, was
Alavi S.A. a DL Framework accuracy proposed employed to extract features from X-ray images.
for COVID-19 score for deep Features were chosen by a feature selection method, i.e.,
Diagnosis from the two- learning analysis of variance (ANOVA), to reduce computations
Chest X-ray class framework and time complexity while overcoming the curse of
Images: classificati demonstrate dimensionality to improve accuracy.
Enhancing on task s
Radiological (COVID-19 exceptional
Diagnosis vs. No- performance
Efficiency findings) in
is 98.72%, diagnosing
and for COVID-19
the from chest
multiclass X-ray
classificati images,
on task achieving
(COVID-19 high
vs. No- accuracy,
findings precision,
vs. recall, and
Pneumoni specificity.
a) it is
92%.
Kaur P.; Development of The DL The The study This work investigates plant diseases in grapevines. Leaf
Harnal an Efficient Deep technique used accuracy demonstrate blight, Black rot, stable, and Black measles are the four
S.; Tiwari Learning Method in the study is score s the types of diseases found in grape plants. Several earlier
R.; for Automatic transfer learning achieved effectivenes research proposals using machine learning algorithms
Upadhya Identification and with the by the s of the were created to detect one or two diseases in grape plant
y S.; Classification of EfficientNet B7 proposed proposed leaves; no one offers a complete detection of all four
Bhatia Grape Plant architecture for method is technique in diseases.
S.; Diseases feature 98.7%. accurately
Mashat extraction from identifying
A.; grape plant leaf and
Alabdali images. classifying
A.M. four types of
diseases in
grape
plants.
Guo P.; Li Enhancing Graph Experimen The novel In this paper, a novel graph convolutional network (GCN)-
L.; Li C.; Preoperative Convolutional ts on MMIF-GCN based mpMRI information fusion module (named MMIF-
Huang Glioma Grading Network (GCN). BraTS202 method GCN) is proposed to comprehensively fuse the tumor
W.; Zhao Accuracy with 0 and improves grading relevant information in mpMRI. Specifically, a
G.; Wang Contextual GliomaHP glioma graph is constructed according to the characteristics of
S.; Wang Fusion of mpMRI PH2018 grading mpMRI data. The vertices are defined as the glioma
M.; Lin Y. Data datasets accuracy by grading features of different slices extracted by the CNN,
show that fusing and the edges reflect the distances between the slices in
MMIF-GCN contextual a 3D volume. The proposed method updates the
effectively information information in each vertex considering the interaction
integrates from mpMRI between adjacent vertices.
grading data,
informatio potentially
n from enhancing
mpMRI clinical
data, decision-
enhancing making for
glioma glioma
grading patients.
performan
ce over
prior
methods.
Najumnis Development of LDBN, PDBN, CNNs The In developing countries like India, Rice is the
sa D.; a Real-Time CNN. achieved proposed fundamental and most important food crop. Also, rice is
Vijayalak System for significant real-time an essential food crop for low and lower middle-income
shmi S.; Detecting ly higher system countries. The disease in the rice leaf affects rice
Paramasi Healthy and accuracy utilizing CNN production which results in severe poverty across the
vam A.; Diseased Rice in for rice leaf country. The proposed work aims to design and build a
Karthikha Leaves to detecting disease real-time system to detect healthy and diseased rice
R.; Enhance rice leaf detection leaves. The reliability analysis of three distinct deep
Mohame Agricultural diseases holds high learning algorithms, namely Linear kernel Deep Belief
d Shuaib Productivity in compared social Networks (LDBN) and Polynomial kernel Deep Belief
Y. Developing to Deep significance, Networks (PDBN), and Convolutional Neural Networks
Countries Belief offering a (CNN) was also performed.
Networks promising
(DBN) solution to
with linear combat rice
and production
polynomia losses and
l kernels. poverty in
developing
countries.
Sakib S.; Remote health model-M1 to
Fouda monitoring has identify
M.M.; Al- become a arrhythmia. We
Mahdawi necessity due to use multiple
M.; reduced publicly available
Mohsen healthcare clinically
A.; access resulting annotated
Oogane from pandemic datasets to
M.; Ando lockdowns and evaluate our
Y.; the increasing proposal. We find
Fadlullah aging population. that denoising by
Z.M. Electrocardiogra model-M1 retains
phy (ECG) is the the features,
standard for which assist the
cardiac model-M2 in
monitoring and achieving high
arrhythmia classification
identification, accuracy,
but it is compared to
inconvenient for using a
long-time remote conventional
monitoring. moving average
Recently, filter. This AI
Magnetocardiogr pipeline
aphy (MCG) architecture is
sensors that promising for
operate at room privacy-
temperature preserving ultra-
became edge medical
available based sensing devices.
on spintronic
sensors.
However, MCG
analysis is
affected by the
low-frequency
noise present at
the
Varzaneh Development of DT, SVM, MLP, K- The The study Predicting severe respiratory failure due to COVID-19 can
Z.A.; a Predictive NN. accuracy emphasizes help triage patients to higher levels of care, resource
Orooji A.; Model for score the allocation and decrease morbidity and mortality. The
Erfannia Assessing achieved developed need for this research derives from the increasing
L.; Intubation Risk in by the predictive demand for innovative technologies to overcome
Shanbeh Hospitalized proposed model's complex data analysis and decision-making tasks in
zadeh M. COVID-19 DT-based potential to critical care units. Hence the aim of our paper is to
Patients Using predictive quantitativel present a new algorithm for selecting the best features
Machine model is y assess from the dataset and developing Machine Learning (ML)
Learning and 93%. illness based models to predict the intubation risk of
Feature Selection severity and hospitalized COVID-19 patients.
Techniques identify
high-risk
COVID-19
 patients,
aiding
frontline
clinicians in
patient
managemen
t and
resource
allocation in
critical care
settings.
Pandey Enhancing CNN CNN. Combining The study The training of deep learning convolutional neural
S.; Computational pipeline emphasizes networks is extremely compute intensive and takes long
Nagwani Performance on and task the times for completion, on all except small datasets. This is
N.K.; Multisystem parallelis importance a major limitation inhibiting the widespread adoption of
Verma S. High- m can of convolutional neural networks in real world applications
Performance boost CNN developing despite their better image classification performance in
Computing performan parallel and comparison with other techniques. Multidirectional
Architectures to ce by up scalable research and development efforts are therefore being
Reduce Training to 18 implementat pursued with the objective of boosting the computational
Times and times on ions of CNNs performance of CNN.
Facilitate Real- multisyste on high-
World Adoption m high- performance
performan computing
ce architecture
computing s to address
architectu the
res, computation
enhancing al intensity
deep and long
learning training
computati times
onal associated
efficiency. with deep
learning
tasks.
Khozeim Development of CNN, AEs The CNN-AE The automation of the diagnosis of this virus is highly
eh F.; CNN-AE for accuracy demonstrate desired. CNN have shown outstanding classification
Sharifrazi Predicting achieved s the performance on image datasets. To date, it appears that
D.; Izadi COVID-19 by CNN-AE potential of COVID computer-aided diagnosis systems based on CNNs
N.H.; Survival Chances is 96.05%. using clinical and clinical information have not yet been analysed or
Joloudari Using Clinical 4o data for explored. We propose a novel method, named the CNN-
J.H.; Data COVID-19 AE, to predict the survival chance of COVID-19 patients
Shoeibi diagnosis, using a CNN trained with clinical information. Notably,
A.; providing a the required resources to prepare CT images are
Alizadehs cost- expensive and limited compared to those required to
ani R.; effective collect clinical data, such as blood pressure, liver disease,
Gorriz alternative etc.
J.M.; to CT image-
Hussain based
S.; Sani methods.
Z.A.;
Moosaei
H.;
Khosravi
A.;
Nahavan
di S.;
Islam
S.M.S.
Loey M.; Classification of GoogleNet, The The study The outcomes show that ResNet18 is the most stable
Mirjalili COVID-19 Cough ResNet18, accuracy demonstrate model to classify the cough sounds from a limited
S. Sounds in Real- ResNet50, achieved s the dataset with a sensitivity of 94.44% and a specificity of
Life ResNet101, by the efficacy of 95.37%. Finally, a comparison of the research with a
Environments MobileNetv2, and model is using deep similar analysis is made. It is observed that the proposed
Using Deep NasNetmobile. 94.9%. learning model is more reliable and accurate than any current
Learning and models for models. Cough research precision is promising enough to
Deep Transfer classifying test the ability for extrapolation and generalization.
Models COVID-19
cough
sounds, with
ResNet18
exhibiting
the most
stable
performance
.
Pandey The CNN training on the basis of
S.; algorithm has the input image.
Nagwani been One hundred
N.K.; implemented for images
Verma S. a central corresponding to
processing unit each class of
based high defects have
performance been used for
multisystem the training in
architecture order to keep the
machine. The training times
multisystem or manageable. The
the hyperparameters
multicomputer is of both the
a parallel convolutional
machine model neural networks
which is were kept
essentially an identical and the
abstraction of programs were
distributed run on the same
memory parallel computational
machines. In cluster to enable
actual practice, fair comparison.
 this model Both the
corresponds to convolutional
high neural network
performance implementations
computing have been
clusters. The observed to train
proposed to nearly 80%
implementation training accuracy
of the in 200 epochs. In
convolutional effect, therefore,
neural network the comparison
training is on the time
algorithm is taken to
based on complete the
modeling the training epochs.
convolutional
neural network
as a
computational
pipeline. The
various functions
or tasks of the
convolutional
neural network
pipeline have
been mapped
onto the multiple
nodes of a
central
processing unit
based high
performance
computing
cluster for task
parallelism. The
pipeline
implementation
provides a first
level
performance
gain through
pipeline
parallelism.
Further
performance
gains are
obtained by
Khan Medical services preserved logs in
A.A.; often involve the a blockchain
Shaikh transmission of distributed
Z.A.; demanding ledger. The
Baitenov information. simulation result
a L.; Thus, there are shows that QoS
Mutaliye stringent is computed on
va L.; requirements for the blockchain
Moiseev se-cure, public network
N.; intelligent, with
Mikhaylo public-network transmission
v A.; quality-of- power = average
Laghari service. This of −10 to −17
A.A.; Idris paper dBm, jitter = 34
S.A.; contributes to ms, delay =
Alshazly three different average of 87 to
H. aspects. First, we 95 ms,
propose a novel throughput =
metaheuristic 185 bytes, duty
approach for cycle = 8%,
medical cost- route of delivery
efficient task and response
schedules, where back variable.
an intelligent Thus, the
scheduler proposed QoS-
manages the ledger is a
tasks, such as potential
the rate of candidate for the
service schedule, computation of
and lists items quality-of-service
utilized by users that is not
during the data limited to e-
processing and healthcare
computation distributed
through the fog applications.
node.
Singh Design and The study Experimen The study The recently discovered coronavirus, SARS-CoV-2, which
P.D.; Kaur Development of utilizes a tal results concludes was detected in Wuhan, China, has spread worldwide and
R.; Singh an Ensemble- synthetic dataset demonstra that the is still being studied at the end of 2019. Detection of
K.D.; Based Classifier of COVID-19 te that the developed COVID-19 at an early stage is essential to provide
Dhiman for Early generated by a proposed ensemble- adequate healthcare to affected patients and protect the
G. Prediction of dataset novel based uninfected community. This paper aims to design and
COVID-19 Cases generation ensemble- classifier, develop a novel ensemble-based classifier to predict
algorithm. based coupled with COVID-19 cases at a very early stage so that appropriate
classifier the convex action can be taken by patients, doctors, health
outperfor hull-based organizations, and the government. In this paper, a
ms approach, synthetic dataset of COVID-19 is generated by a dataset
popular offers a generation algorithm. A novel ensemble-based classifier
classifiers. more of machine learning is employed on the COVID-19
accurate dataset to predict the disease. A convex hull-based
and efficient approach is also applied to the data to improve the
method for proposed novel, ensemble-based classifier’s accuracy
predicting and speed.
 COVID-19
cases at an
early stage.
Miao R.; Overcoming The study The UMLF- The study With the rapid spread of COVID-19 worldwide, quick
Dong X.; Limitations in introduces an COVID concludes screening for possible COVID-19 patients has become the
Xie S.-L.; Deep Learning unsupervised model that the focus of international researchers. Recently, many deep
Liang Y.; Models for Fast meta-learning achieves a UMLF-COVID learning-based Computed Tomography (CT) image/X-ray
Lo S.-L. COVID-19 model, UMLF- 3-10% model image fast screening models for potential COVID-19
Screening Using COVID, for fast improvem effectively patients have been proposed. However, the existing
CT and X-Ray screening of ent in complement models still have two main problems. First, most of the
Images COVID-19 accuracy s current existing supervised models are based on pre-trained
patients using CT compared rapid model parameters. The pre-training model needs to be
and X-ray to existing screening constructed on a dataset with features similar to those in
images. models. models for COVID-19 X-ray images, which limits the construction and
COVID-19 use of the model. Second, the number of categories
patients based on the X-ray dataset of COVID-19 and other
using CT and pneumonia patients is usually imbalanced. In addition,
X-ray the quality is difficult to distinguish, leading to non-ideal
images. results with the existing model in the multi-class
classification COVID-19 recognition task.
Zheng Addressing The study The The study The sudden increase in coronavirus disease 2019
W.; Yan Challenges in proposes a multi- proposed concludes (COVID-19) cases puts high pressure on healthcare
L.; Gou COVID-19 modal approach that services worldwide. At this stage, fast, accurate, and
C.; Diagnosis: knowledge graph achieves integrating early clinical assessment of the disease severity is vital.
Zhang Integrating attention significant textual In general, there are two issues to overcome: (1) Current
Z.-C.; Textual and embedding ly information DL-based works suffer from multimodal data adequacy
Jason Image approach for improved from doctor- issues; (2) In this scenario, multimodal (e.g., text, image)
Zhang J.; Information for COVID-19 classificati patient information should be taken into account together to
Hu M.; Accurate diagnosis. on dialogues make accurate inferences. To address these challenges,
Wang F.- Inference Using performan with we propose a multi-modal knowledge graph attention
Y. Multimodal Data ce, correspondin embedding for COVID-19 diagnosis. Our method not only
demonstra g medical learns the relational embedding from nodes in a
ting images constituted knowledge graph but also has access to
accuracy enhances medical knowledge, aiming at improving the
gains the accuracy performance of the classifier through the mechanism of
across and medical knowledge attention.
different effectivenes
modalities s of COVID-
of multi- 19
modal diagnosis.
data for
COVID-19
diagnosis.
Al-ali A.; Development of Adaptive Neuro- The The study Among the most leading causes of mortality across the
Elharrous an Auto- Fuzzy Inference results concludes globe are infectious diseases which have cost
s O.; Diagnosis System (ANFIS), show that the tremendous lives with the latest being coronavirus
Qidwai System for gray level co- promising ANFIS-based (COVID-19) that has become the most recent challenging
U.; Al- COVID-19 occurrence accuracy method is issue. The extreme nature of this infectious virus and its
Maaddee Detection from matrix (GLCM) for the effective for ability to spread without control has made it mandatory
d S. Chest X-ray ANFIS- automaticall to find an efficient auto-diagnosis system to assist the
 Images Using based y detecting people who work in touch with the patients.
ANFIS and GLCM method. It COVID-19
Texture Analysis performs from chest
comparabl X-ray
y to deep images,
learning especially in
technique situations
s, even with limited
without dataset
large sizes.
datasets.
Oza M.; Development of The study The DL- The It has been observed that in the present situation of a
Rani G.; a Deep Learning- utilizes a based proposed pandemic the misleading news specifically about COVID-
Dhaka Based Tool for multiclass classifier DL-based 19 are increasing day by day. Multiple research groups
V.S.; Classifying classifier based achieved tool identified this challenge and proposed machine learning-
Patel H. COVID-19 News on deep learning an successfully based binary classifiers for categorizing news articles
Articles into techniques. accuracy addresses into fake and true classes. But, none of them addressed
Fake, True, and of 88%. the the challenge of identifying the misleading news. Also,
Misleading challenge of the existing research works do not focus on examining
Categories identifying the reliability of the news about COVID-1. Moreover,
misleading there is a lack of complete system that provide an
news, a gap integration of front-end and back-end where the users
not covered can check the reliability of news article and get
by existing recommendation of the sources for validation of news
binary articles.
classifiers
focused
solely on
fake and
true news.
Pandey Enhancing CNN The study The The The study's findings indicate that integrating pipeline
S.; Performance utilizes CNNs performan proposed parallelism with task parallelism can substantially boost
Nagwani through Pipeline ce pipeline CNN training efficiency, achieving an 18-fold
N.K.; Parallelism for analysis of computation performance gain. This advancement has the potential to
Verma S. Improved the model for mitigate the computational resource constraints and time
Computational pipeline CNNs demands that currently limit the widespread adoption of
Efficiency model, successfully CNNs, promoting their use in new applications.
3.5 based on addresses
computati the
onal challenge of
experimen extensive
ts, shows training
a net times by
performan enhancing
ce gain of performance
18X. through
parallelism
techniques.
Abdulkar Leveraging AI CNN, RNN, GAN. AI The study Recent advancements, particularly in AI tools, have been
eem M.; Tools to Combat systems suggests investigated to monitor the spread of the coronavirus,
Petersen COVID-19: have forming identify high-risk patients, and diagnose the disease. This
S.E. Detection, shown specialized paper explores AI applications in detection, diagnosis,
Diagnosis, potential teams with epidemiological predictions, forecasting, and social
Predictions, and in expertise in control for COVID-19, highlighting successes and
Social Control diagnosin medical data challenges. It also reviews AI systems developed for
g COVID- privacy, the diagnosing COVID-19 using chest CT and X-ray images.
19, use of
predicting federated
mortality learning,
rates, and the
long-term utilization of
hospitaliza biobank
tion, and biomedical
patient data.
outcomes.
Ming Y.; Dynamic The study The study The results In this article with the aid of adaptively tuned deep
Zandi Y.; Stability Analysis utilizes a deep showcases suggest that neural network (DNN), dynamic stability analysis of the
Gholizad of Sandwich neural network the DNN- the sandwich structure has been investigated. Due to finding
eh M.; Structures Using (DNN). based momentum- the design-points features, the numerical solution
Oslub K.; Adaptively Tuned model's based procedure called two-dimensional generalized differential
Khadimal Deep Neural effectiven optimizer quadrature technique has been applied to the ordinary
lah M.A.; Networks ess in coupled with differential equations of the current structure system
Issakhov accurately the acquired on the foundation of the kinematic theory with
A. analyzing adaptively refined higher order terms.
the tuned DNN
stability of is a
sandwich promising
structures. tool for
future
research in
stability
analysis of
various
structures.
Guo X.; Predicting The study The study The study The presented paper is the first attempt to apply deep
Liu Y.; Frequency utilizes a fully demonstra introduces a learning for predicting the frequency characteristics of a
Wang G. Characteristics of connected neural tes the novel use of magneto-electro-elastic (MEE) annular nano/microdisk
MEE Annular network (FCNN). effectiven deep (MEEAD). The optimum amount of the factors
Nano/Microdisks ess of the learning to participating in the mechanism the fully connected
Using Deep FCNN- predict the neural network are achieved through the optimizer based
Learning and based frequency on the momentum. The side of the mentioned approach
Momentum- approach characteristi employed in this investigation would be due to its high
Based in cs of MEEAD accuracy along with lower epochs required for training
Optimization accurately with high the multi-layered network.
predicting accuracy
the and
frequency efficiency.
behavior
of MEEAD.
Regin R.; Developing an CNN. The The study Finally, we analyze convex hull, Naïve bayes and CNN
Rajest Efficient analysis of presents an algorithms using real-world datasets to identify and
S.S.; Distributed Fault the effective organize the faults. Simulation and experimental
Singh B. Detection Convex approach for outcomes retain feasibility and efficiency and show that
Method for Hull, fault the CNN algorithm has better-identified faults than the
Wireless Sensor Naïve detection in convex hull algorithm based on performance metrics.
Networks Bayes, WSNs by
and CNN integrating
algorithms the Convex
using real- Hull
world algorithm,
datasets Naïve Bayes
demonstra classifier,
tes that and CNN.
the CNN
algorithm
outperfor
ms the
Convex
Hull
algorithm
in
identifying
faults
based on
performan
ce
metrics.
Chen Y.; Unsupervised CNN. The The study White matter fiber clustering (WMFC) enables
Zhang Deep Learning results presents an parcellation of white matter tractography for applications
C.; Song Framework for demonstra effective such as disease classification and anatomical tract
Y.; Makris White Matter te that the WMFC segmentation. However, the lack of ground truth and the
N.; Rathi Fiber Clustering. proposed framework ambiguity of fiber data (the points along a fiber can
Y.; Cai framework based on equivalently be represented in forward or reverse order)
W.; achieves unsupervise pose challenges to this task. We propose a novel WMFC
Zhang F.; superior d deep framework based on unsupervised DL
O’Donnel performan learning. By
l L.J. ce and leveraging
efficiency self-
in WMFC. supervised
learning and
incorporatin
g anatomical
segmentatio
n data, the
framework
improves
clustering
performance
and
anatomical
 coherence.
Abbas S.; Advanced Extremely The The model One in eight women all over the world is affected by this
Jalil Z.; Predictive Model Randomized Tree results outperforms disease. Almost half a million women annually do not
Javed for Breast Cancer algorithm. show that existing survive this fight and die from breast cancer. Machine
A.R.; Diagnosis: the BCD- algorithms, learning algorithms have proven to outperform all
Batool I.; Enhancing WERT achieving existing solutions for the prediction of breast cancer
Khan Feature Selection model the highest using models built on the previously available data. In
M.Z.; and outperfor accuracy this paper, a novel approach named BCD-WERT is
Noorwali Classification ms SVM, and proposed that utilizes the Extremely Randomized Tree
A.; Accuracy Using Random highlighting and Whale Optimization Algorithm (WOA) for efficient
Gadekall Machine Forest, the feature selection and classification. WOA reduces the
u T.R.; Learning Kernel importance dimensionality of the dataset and extracts the relevant
Akbar A. Algorithms SVM, of efficient features for accurate classification.
Decision feature
Tree, selection.
Logistic
Regressio
n,
Stochastic
Gradient
Descent,
Gaussian
Naive
Bayes,
and k-
Nearest
Neighbor.
Dutta Developing an Convolutional The The study The major challenge is to cope with the mobility issue of
Pramanik Effective feature results concludes SMDs. User’s unpredicted mobility makes the SMDs really
P.K.; Prediction Model extraction, LSTM, show that that the unstable resources. Selecting such erratic resources for
Sinhabab for User GRU. the convolutiona job schedule would result in frequent job offloading and,
u N.; Availability in proposed l LSTM in the worst case, job loss, which would affect the overall
Kwak K.- Mobile Cloud convolutio model is performance and the quality of service of MCC. In a Local
S.; Computing to nal LSTM highly MCC, generally, a set of users are available for a certain
Choudhu Enhance Job model effective for period regularly. Based on this information, the chances
ry P. Scheduling and outperfor predicting of a user being available for a certain duration from a
Minimize ms basic the given point of time can be predicted.
Offloading Loss. LSTM, availability
GRU, and of users in
ARIMA an MCC
models in environment
predicting , thereby
user reducing
availability unnecessary
. job
offloading
and job loss.
Rao S.; Development of Fully connected The The study Additionally, we investigate prospects of efficient
Narayan Non-Invasive neural networks customize concludes implementation by examining the impact on the
aswamy COVID-19 (FCNN), CNN. d VGG-13 that the detection performance by pruning the fully connected
V.; Detection Tools model VGG-13 neural network based on the Lottery Ticket Hypothesis
Esposito Using Machine achieved model, with (LTH) optimization process. In general, pruned neural
M.; Learning and an data networks have been shown to provide similar
Thiagaraj Deep Learning accuracy augmentatio performance gains to that of unpruned networks with
an J.J.; on Coughing of 82.23% n and reduced computational complexity in a variety of signal
Spanias Sound Features. AUROC on specialized processing applications
A. the loss
validation functions, is
set and highly
78.3% effective for
AUROC on automated
the test COVID-19
set, with a diagnosis
sensitivity using
of coughing
80.49%. sounds,
outperformi
ng other
neural
network
models
tested.
Mishra Enhancing Data CNN, RNN. The The study CNN and RNNs have the ability to find the accurate result
V.K.; Processing proposed concludes in images and text respectively. The best classification
Kumar V.; Efficiency in framework that results are still awaited due to the high cost of
Pandey Image and Text achieves integrating computation and high memory requirements of CNN and
N.K. Classification improved feedback RNN. Our work suggests a framework that improves the
Tasks Using classificati mechanisms quality of data at various layers by providing feedback to
Optimized on results into CNN suggested system.
Convolutional for image and RNN
and Recurrent and text architecture
Neural Networks tasks by s
addressin significantly
g enhances
computati data
onal costs processing
and quality,
memory improving
requireme accuracy in
nts, image and
leading to text
error-free classification
processing tasks.
systems.
Valizadeh The study aims FCN, CRF, CNN. The study The study The presented aspects, including all-convolution network,
A.; to compare the offers a concludes sampling methods, FCN connector with CRF methods,
Shariatee traditional thorough that deep extended CNN methods, improvements in network
M. method of review of neural structure, pyramid methods, multistage and multifeature
semantic deep network- methods, supervised methods, semiregulatory methods,
segmentation neural based and nonregulatory methods, are then thoroughly
 with the deep network- methods explored in current methods based on the deep neural
neural network based have shown network.
method. It methods significant
reviews the for advancemen
traditional semantic ts in
segmentation segmentat semantic
methods and ion, segmentatio
datasets, contrastin n, offering
explores various g them improved
aspects of deep with accuracy
neural network- traditional and
based methods approache efficiency
for semantic s to compared to
segmentation, assess traditional
and draws effectiven methods.
conclusions on ess and
the applicabili
advancements in ty in the
this field. field.
Rashid A Social Sensing- The study The Model CovidSens Specifically, the vision of CovidSens attempts to answer
M.T.; Based Risk Alert primarily focuses introduces represents a the questions: How to distill reliable information about
Wang D. System for Real- on leveraging CovidSens promising the COVID-19 with the coexistence of prevailing rumors
Time Monitoring social sensing for real- approach to and misinformation in the social media? How to inform
of COVID-19 techniques to time harnessing the general public about the latest state of the spread
Propagation extract and COVID-19 social media timely and effectively, and alert them to remain
Using Online analyze data monitorin data for prepared? How to leverage the computational power on
Social Media from online g through monitoring the edge devices (e.g., smartphones, IoT devices, UAVs)
Data social media social and alerting to construct fully integrated edge-based social sensing
platforms, rather sensing, the public platforms for rapid detection of the COVID-19 spread? In
than specific highlightin about the this vision paper, we discuss the roles of CovidSens and
deep learning g spread of identify the potential challenges in developing reliable
techniques. challenges COVID-19. social sensing-based risk alert systems.
and
advocatin
g
interdiscip
linary
collaborati
on for
robust risk
alert
systems.
Majedka The study aims The study The Queuing This Queuing theory analysis has been used in hospitals
n N.A.; to analyze employs queuing average theory helps and other healthcare settings; its use in this sector is not
Idrees healthcare theory, a waiting healthcare widespread. Consequently, the queue waiting line is an
B.A.; queue waiting mathematical time for staff predict effective scientific tool in the performance of waiting
Ahmed times in Duhok approach, to each daily queue lines analysis. The main parameters used to measure the
O.M.; city using analyze waiting citizen lengths and performance of the systems are the length of the queue
Haji L.M.; queuing theory times in checked potential line, utilization of the server, and the delays for arrivals.
Dino H.I. to enhance healthcare by the client
 patient settings and healthcare dropout
satisfaction and predict queue team in rates, aiding
reduce the risk lengths and the queue in better
of epidemics waiting times. is 32.44 resource
such as COVID- minutes. managemen
19. t and
enhancing
patient
satisfaction
in Duhok
city.
Montecin Proposing a The study The The study This search space allows CNN architectures to have any
o D.A.; Novel Two-Level utilizes proposed demonstrate sizes, shapes, and skip-connections among nodes. The
Perez Genetic neuroevolution, two-level s the two-level GA was applied to the pattern recognition
C.A.; Algorithm to specifically a genetic effectivenes problem on seven datasets, five MNIST-Variants, Fashion-
Bowyer Address Fidelity- two-level genetic algorithm s of the two- MNIST, and CIFAR-10, achieving significantly better
K.W. Time Trade-Off in algorithm, to achieves level GA in results than all those previously published. Our results
Neuroevolution automatically error optimizing show an improvement of 39.89% (4.2% error reduction)
for CNN find optimal CNN reductions CNN on the most complex dataset of MNIST (MRDBI), and on
Architectures architectures for ranging architecture average 30.52% (1.35% error reduction) on all the five
pattern from s, achieving datasets. Furthermore, we show that our algorithm
recognition tasks 1.35% to improved performed as well as precise-training GA, but took only
on various 4.2% on performance the time of a fast-training GA.
datasets. seven on pattern
datasets, recognition
including tasks while
MNIST- maintaining
Variants, computation
Fashion- al efficiency.
MNIST,
and CIFAR-
10.
Wen A.; DL-Based Autoencoders. The study The study Coronavirus Disease 2019 has emerged as a significant
Wang L.; Syndromic shows highlights global concern, triggering harsh public health restrictions
He H.; Surveillance for that the potential in a successful bid to curb its exponential growth. As
Liu S.; Fu Early Detection autoencod of DL-based discussion shifts towards relaxation of these restrictions,
S.; Sohn of COVID-19 er-based syndromic there is significant concern of second-wave resurgence.
S.; Kugel Outbreaks and methods surveillance
J.A.; Differentiation effectively for early
Kaggal from Influenza- detect detection of
V.C.; Like Illnesses influenza COVID-19
Huang outbreaks outbreaks
M.; Wang and and
Y.; Shen differentia differentiatio
F.; Fan J.; te them n from
Liu H. from other similar
influenza- illnesses.
like
illnesses,
such as
 the
common
cold.
D. Comprehensive DL technique for The study The study Social network based applications have experienced
Camacho Review and New analyzing defines contributes exponential growth in recent years. One of the reasons
, Á. Metrics Proposal structured data four to the for this rise is that this application domain offers a
Panizo- in Social Network like images is dimension understandi particularly fertile place to test and develop the most
LLedot, Analysis: CNN. s in SNA— ng and advanced computational techniques to extract valuable
G. Bello- Quantitative Pattern & evaluation information from the Web. The main contribution of this
Orgaz, A. Analysis of Tools Knowledg of social work is three-fold: (1) we provide an up-to-date literature
Gonzalez and Frameworks e network review of the state of the art on social network analysis
-Pardo, E. discovery, technologies (SNA); (2) we propose a set of new metrics based on four
Cambria Informatio by providing essential features (or dimensions) in SNA; (3) finally, we
n Fusion & a provide a quantitative analysis of a set of popular SNA
Integratio comprehensi tools and frameworks.
n, ve review,
Scalability proposing
, and new metrics,
Visualizati and
on—and analyzing
proposes popular SNA
new tools and
metrics frameworks.
(degrees)
within
these
dimension
s to
evaluate
SNA
software
tools and
framework
s.
Shi B.; Ye Early Detection Although not The study The study The experimental validation confirms the efficacy of
H.; and Severity explicitly stated, demonstra underscores ESMA in benchmark function testing and its superiority
Zheng J.; Discrimination of traditional tes the the potential over other optimization algorithms. The strong
Zhu Y.; COVID-19 Using machine learning effectiven effectivenes performance of ESMA-SVM in discriminating COVID-19
Heidari Clinical Data: A techniques are ess of the s of the severity underscores its potential for practical application
A.A.; Robust predominantly proposed ESMA-SVM in medical settings, contributing to early intervention and
Zheng L.; Methodology utilized in this methodolo methodolog optimized resource allocation in the management of the
Chen H.; Incorporating study, with a gy, ESMA- y for early pandemic.
Wang L.; Patient focus on support SVM, in detection
Wu P. Information and vector machine early and severity
Hematological (SVM) modeling. recognitio discriminatio
Indexes n and n of COVID-
discrimina 19 using
tion of clinical data.
COVID-19
severity
 using
clinical
data.
Tassone The goal of this The primary DL The initial In However, CNN-based classifiers demonstrate significant
J.; Yan P.; study is to technique results conclusion, improvement in predictive accuracy. The models trained
Simpson develop a utilized in this from SVM the study with manually labeled samples and synthetically
M.; predictive model study was CNN. and finds that generated tweets achieve higher accuracy scores and
Mendhe for identifying XGBoost predictive area under the curve (AUC) values, indicating superior
C.; Mago positively models analysis performance in identifying drug-related tweets.
V.; referenced, drug- show using CNN- Association rule mining, when combined with CNN-based
Choudhu related tweets relatively based classifiers, further enhances the predictive capability by
ry S. on Twitter, using low classifiers identifying keywords strongly associated with drug-
social media accuracy shows positive classifications, such as "smoke," "cocaine," and
data analysis. scores and promise in "marijuana."
discrimina accurately
tion identifying
capability positively
for referenced,
identifying drug-related
positively tweets on
reference Twitter.
d, drug-
related
tweets.
Alomari Proposing a The study The In The proposed software tool, leveraging unsupervised
E.; Katib Software Tool for primarily utilizes study's conclusion, machine learning techniques such as Latent Dirichlet
I.; Enhancing unsupervised software the study Allocation (LDA), enables the detection of government
Albeshri COVID-19- machine learning tool highlights pandemic measures, public concerns, and macro-
A.; Related Studies: techniques, successful the concerns related to COVID-19 in the Kingdom of Saudi
Mehmoo Unsupervised including Latent ly detects significance Arabia. The study demonstrates the tool's effectiveness
d R. Latent Dirichlet Dirichlet governme of utilizing in identifying temporal and spatio-temporal relationships
Allocation (LDA) Allocation (LDA). nt social media between these phenomena, providing valuable insights
for Analyzing pandemic analytics, into the progression of events and government responses
Arabic Twitter measures particularly during the pandemic.
Data and public Twitter data,
concerns for
from understandi
Arabic ng the
Twitter dynamics of
data the COVID-
during the 19
COVID-19 pandemic.
pandemic
in the
Kingdom
of Saudi
Arabia
(KSA).
Tahir H.; Exploring Two The study The study In The study demonstrates that using ResNet-101 for
Iftikhar Dimensions of primarily achieves conclusion, patient registration slips and Faster R-CNN, Mask-CNN,
A.; COVID-19 employs deep an 82% the study and ResNet-50 for chest X-rays achieves high forecasting
Mumraiz Prediction: learning forecastin demonstrate accuracy in COVID-19 prediction tasks. It emphasizes the
M. ResNet-101 techniques, g s the critical role of advanced machine learning techniques in
Application to specifically accuracy potential of accurate and efficient disease forecasting, crucial for
Patient ResNet-101, to for COVID- deep early detection, prevention, and management of COVID-
Registration Slips forecast COVID- 19 learning 19 outbreaks. Furthermore, the study underscores the
and Chest X- 19 positive cases positive techniques need for continued research and optimization of deep
rays. using patient cases for learning models to enhance COVID-19 prediction
registration slips. using predicting capabilities.
ResNet- COVID-19 in
101. two
dimensions:
utilizing
patient
registration
slips and
chest X-rays.
Sarv Comparative CNN is employed The The study Moreover, testing on diverse datasets not used in the
Ahrabi Analysis of DL for the binary proposed demonstrate training process yields an accuracy of up to 93%, further
S.; Techniques for classification of CNN s the validating the robustness of the proposed approach.
Scarpiniti COVID-19 COVID-19 variant effectivenes
M.; Detection in X- infected X-ray achieves s of the
Baccarell ray Images: images. excellent proposed
i E.; Proposing an performan CNN variant
Momenz Optimized CNN ce with an in accurately
adeh A. Variant. overall detecting
accuracy COVID-19
of 99.8% from X-ray
on a images.
recent
dataset of
COVID-19
X-rays,
showcasin
g its
effectiven
ess in
detecting
COVID-19.
Al-Andoli Systematic Deep Neural The study In In addition, the relevant studies on DL with parallel and
M.N.; Tan Review and Networks (CNN). reviews conclusion, MH approaches are reviewed and their implications on DL
S.C.; Future Directions Communit the study models are highlighted to prospect effective solutions to
Cheah in Community y emphasizes overcome the challenges of DL-based CD methods.
W.P.; Tan Detection Detection the ongoing
S.Y. Methods: From (CD) effort to
Conventional methods, develop
Machine including precise and
Learning to Deep conventio efficient
Learning and nal deep
Meta-Heuristic machine learning-
 Algorithms learning based
and deep methods for
learning- Community
based Detection
approache (CD),
s, focusing especially
on tailored for
challenges large
with large Complex
Complex Networks
Networks (CNs).
(CNs).
Rahman B5G Framework Semantic The The However, existing DL algorithms suffer from two crucial
M.A.; for Global Augmentation, implement proposed drawbacks: first, the training requires a large COVID-19
Hossain COVID-19 Visualization ed B5G framework dataset on various dimensions, which is difficult for any
M.S.; Diagnosis Using framework addresses local authority to manage. Second, the DL results require
Alrajeh Edge Computing for COVID- the ethical approval and explanations from healthcare
N.A.; and Deep 19 challenges providers and other stakeholders in order to be accepted.
Guizani Learning. diagnosis of managing In this article, we propose a B5G framework that supports
N. shows large COVID- COVID-19 diagnosis, leveraging the low-latency, high-
promising 19 datasets bandwidth features of the 5G network at the edge. Our
test and gaining framework employs a distributed DL paradigm where
results, acceptance each COVID-19 edge employs its own local DL framework
leveraging from and uses a three-phase reconciliation with the global DL
the low- healthcare framework.
latency providers
and high- and
bandwidth stakeholders
features .
of the 5G
network at
the edge.
Ref Study Goal DL Dataset Optimizatio Results Deductions Others
Technique n Algorithm
Used

Shekar P.R. Rainfall- ANN, KNN, CNN-RNN CNN-RNN Calibration

et al. Runoff XGBoost, model model period:
(2024) Modeling RF, CNN, showed could be 2003–
CNN-RNN superior valuable 2007;
performanc for validation
e in sustainable period:
simulating water 2008–2009
monthly resource
rainfall- manageme
runoff nt, flood
control,
and
environme
ntal
planning

Issahaku F.- COVID-19 Modified CXR High Our Dataset:

L.Y. et al. Detection VGG16, images, efficacy in method 2059
(2024) Faster R- cough detecting excelled, COVID-19
CNN recordings COVID-19 surpassing positive
 existing CXR
networks images,
25000
cough
recordings

A.A. et al. Intrusion CGAN, High Combining Study

(2024) Detection VGG19, AO accuracy in Game focuses on
System for detecting Theory, effective
Smart cyber- swarm protection
Grids attacks intelligence against
with , and DL cyber
99.82% techniques threats in
accuracy provides a smart grids
robust
solution for
SG
intrusion
detection

A. E. et al. Parking ANFIS-RNN, Superior Significant ANFIS-RNN

(2024) Occupancy ANFIS- performanc improveme improved
Forecasting LSTM, e nts in MSE by
ANFIS-GRU compared parking 30.61%,
to occupancy RMSE by
standalone forecasting 16.70%,
DL models MAE by
21.21%,
MAPE by
21.58%, R2
 by 1.03%

J.S. et al. Sentiment Capsule- High CRDC Accuracy:

(2024) Analysis based RNN, accuracy model IMDB
CNN, CRDC with the enhances (88.15%),
model CRDC sentiment Toxic
model analysis (98.28%),
outperform accuracy CrowdFlow
ing others significantl er
y (92.34%),
ER
(95.48%)

T. et al. Lung CNNs Medical Achieved a Effective in

(2024) Cancer Imaging remarkable diagnosing
Identificati accuracy and
on and rate of treating
Classificati 0.994 lung
on cancer
through DL
methods