AIMS Mathematics,
Journal Year:
2024,
Volume and Issue:
9(12), P. 35678 - 35701
Published: Jan. 1, 2024
<p>Autonomous
vehicles
(AVs),
particularly
self-driving
cars,
have
produced
a
large
amount
of
interest
in
artificial
intelligence
(AI),
intelligent
transportation,
and
computer
vision.
Tracing
detecting
numerous
targets
real-time,
mainly
city
arrangements
adversarial
environmental
conditions,
has
become
significant
challenge
for
AVs.
The
effectiveness
vehicle
detection
been
measured
as
crucial
stage
visual
surveillance
or
traffic
monitoring.
After
developing
driver
assistance
AV
methods,
weather
conditions
an
essential
problem.
Nowadays,
deep
learning
(DL)
machine
(ML)
models
are
critical
to
enhancing
object
AVs,
conditions.
However,
according
statistical
learning,
conventional
AI
is
fundamental,
facing
restrictions
due
manual
feature
engineering
restricted
flexibility
adaptive
environments.
This
study
presents
the
explainable
with
fusion-based
transfer
on
adverse
autonomous
(XAIFTL-AWCDAV)
method.
XAIFTL-AWCDAV
model's
main
aim
detect
classify
AVs
challenging
scenarios.
In
preprocessing
stage,
model
utilizes
non-local
mean
filtering
(NLM)
method
noise
reduction.
Besides,
performs
extraction
by
fusing
three
models:
EfficientNet,
SqueezeNet,
MobileNetv2.
denoising
autoencoder
(DAE)
technique
employed
Next,
DAE
method's
hyperparameter
selection
uses
Levy
sooty
tern
optimization
(LSTO)
approach.
Finally,
ensure
transparency
predictions,
integrates
(XAI)
techniques,
utilizing
SHAP
visualize
interpret
each
feature's
impact
decision-making
process.
efficiency
validated
comprehensive
studies
using
benchmark
dataset.
Numerical
results
show
that
obtained
superior
value
98.90%
over
recent
techniques.</p>
Engineering Technology & Applied Science Research,
Journal Year:
2025,
Volume and Issue:
15(1), P. 20529 - 20537
Published: Feb. 2, 2025
The
need
to
develop
ecologically
friendly
sustainable
building
materials
is
made
apparent
by
the
worldwide
construction
industry's
substantial
contribution
global
greenhouse
gas
emissions.
use
of
supplemental
in
concrete
one
potential
solution
lessen
environmental
footprint.
Thus,
purpose
this
work
Machine
Learning
(ML)
algorithms
forecast
and
create
an
empirical
formula
for
Compressive
Strength
(CS)
with
materials.
Six
distinct
ML
models—XGBoost,
Linear
Regression,
Decision
Tree,
k-Nearest
Neighbors,
Bagging,
Adaptive
Boosting—were
trained
tested
using
a
dataset
that
included
359
experimental
data
varying
mix
proportions.
most
significant
factors
used
as
input
parameters
are
cement,
aggregates,
water,
superplasticizer,
silica
fume,
ambient
curing,
material.
Several
statistical
measures,
such
Mean
Absolute
Error
(MAE),
coefficient
determination
(R2),
Square
(MSE),
were
evaluate
models.
XGBoost
model
outperformed
other
models
R2
values
0.99
at
training
stage.
To
ascertain
how
affected
outcome,
feature
importance
analysis
Shapely
Additive
exPlanations
(SHAP)
was
conducted.
It
demonstrated
curing
age
cement
type
significantly
strength
high
SHAP
values.
By
eliminating
procedures,
reducing
demand
labor
resources,
increasing
time
efficiency,
offering
insightful
information
enhancing
manufacturing
concrete,
research
advances
low-cost
production
USA
industry.
PLoS ONE,
Journal Year:
2025,
Volume and Issue:
20(3), P. e0316287 - e0316287
Published: March 6, 2025
The
accurate
prediction
and
interpretation
of
corporate
Environmental,
Social,
Governance
(ESG)
greenwashing
behavior
is
crucial
for
enhancing
information
transparency
improving
regulatory
effectiveness.
This
paper
addresses
the
limitations
in
hyperparameter
optimization
interpretability
existing
models
by
introducing
an
optimized
machine
learning
framework.
framework
integrates
Improved
Hunter-Prey
Optimization
(IHPO)
algorithm,
eXtreme
Gradient
Boosting
(XGBoost)
model,
SHapley
Additive
exPlanations
(SHAP)
theory
to
predict
interpret
ESG
behavior.
Initially,
a
comprehensive
dataset
was
developed
through
extensive
literature
review
expert
interviews.
IHPO
algorithm
then
employed
optimize
hyperparameters
XGBoost
forming
IHPO-XGBoost
ensemble
model
predicting
Finally,
SHAP
used
model's
outcomes.
results
demonstrate
that
achieves
outstanding
performance
greenwashing,
with
R²,
RMSE,
MAE,
adjusted
R²
values
0.9790,
0.1376,
0.1000,
0.9785,
respectively.
Compared
traditional
HPO-XGBoost
combined
other
algorithms,
exhibits
superior
overall
performance.
analysis
using
highlights
key
features
influencing
outcomes,
revealing
specific
contributions
feature
interactions
impacts
individual
sample
features.
findings
provide
valuable
insights
regulators
investors
more
effectively
identify
assess
potential
behavior,
thereby
efficiency
investment
decision-making.
With
increased
competition
in
the
supermarket
industry,
there
is
an
need
for
higher-order
predictive
analytics
to
garner
insight
into
consumer
behavior
optimal
sales
strategies.
Therefore,
this
research
has
presented
a
predictionusing
ensemble
machine
learning
approach
by
considering
multiple
algorithms:
Random
Forest,
XGBoost,
and
Support
Vector
Machine,
which
further
improve
accuracy
avoid
possible
overfitting.
This
paper
comprehensive
data
preprocessing
feature
engineering,
with
implementation
of
stacking
model,
resulted
excellent
performance.
The
model
achieved
best
R²
value
0.9990and
least
mean
absolute
error.
results
showed
that
techniques
are
very
promising
prediction
provide
powerful
tool
supermarkets
making
better
decisions,
optimizing
inventories,
conducting
focused
marketing.
Hybrid
models
should
be
explored
future
research,
addition
more
external
factors
accuracy.
