Heliyon,
Год журнала:
2024,
Номер
10(12), С. e32400 - e32400
Опубликована: Июнь 1, 2024
Pests
are
a
significant
challenge
in
paddy
cultivation,
resulting
global
loss
of
approximately
20%
rice
yield.
Early
detection
insects
can
help
to
save
these
potential
losses.
Several
ways
have
been
suggested
for
identifying
and
categorizing
fields,
employing
range
advanced,
noninvasive,
portable
technologies.
However,
none
systems
successfully
incorporated
feature
optimization
techniques
with
Deep
Learning
Machine
Learning.
Hence,
the
current
research
provided
framework
utilizing
detect
categorize
photos
promptly.
Initially,
will
gather
image
dataset
it
into
two
groups:
one
without
other
insects.
Furthermore,
various
pre-processing
techniques,
such
as
augmentation
picture
filtering,
be
applied
enhance
quality
eliminate
any
unwanted
noise.
To
determine
analyze
deep
characteristics
an
image,
architecture
incorporate
5
pre-trained
Convolutional
Neural
Network
models.
Following
that,
selection
including
Principal
Component
Analysis
(PCA),
Recursive
Feature
Elimination
(RFE),
Linear
Discriminant
(LDA),
tool
called
Lion
Optimization,
were
utilized
order
further
reduce
redundant
number
features
that
collected
study.
Subsequently,
process
carried
out
by
7
ML
algorithms.
Finally,
set
experimental
data
analyses
has
conducted
achieve
objectives,
proposed
approach
demonstrates
Extracted
Vectors
ResNet50
Logistic
Regression
PCA
achieved
highest
accuracy,
precisely
99.28%.
present
idea
significantly
impact
how
diagnosed
field.
Expert Systems with Applications,
Год журнала:
2024,
Номер
255, С. 124777 - 124777
Опубликована: Июль 14, 2024
Accurately
estimating
the
unknown
parameters
of
photovoltaic
(PV)
models
based
on
measured
voltage-current
data
is
a
challenging
optimization
problem
due
to
its
high
nonlinearity
and
multimodality.
An
accurate
solution
this
essential
for
efficiently
simulating,
controlling,
evaluating
PV
systems.
There
are
three
different
models,
including
single-diode
model,
double-diode
triple-diode
with
five,
seven,
nine
parameters,
respectively,
proposed
represent
electrical
characteristics
systems
varying
levels
complexity
accuracy.
In
literature,
several
deterministic
metaheuristic
algorithms
have
been
used
accurately
solve
hard
problem.
However,
problem,
methods
could
not
achieve
solutions.
On
other
side,
algorithms,
also
known
as
gradient-free
methods,
somewhat
good
solutions
but
they
still
need
further
improvements
strengthen
their
performance
against
stuck-in
local
optima
slow
convergence
speed
problems.
Over
last
two
years,
recent
better
improve
avoid
tackle
continuous
majority
those
has
investigated.
Therefore,
in
paper,
nineteen
recently
published
such
Mantis
search
algorithm
(MSA),
spider
wasp
optimizer
(SWO),
light
spectrum
(LSO),
growth
(GO),
walrus
(WAOA),
hippopotamus
(HOA),
black-winged
kite
(BKA),
quadratic
interpolation
(QIO),
sinh
cosh
(SCHA),
exponential
distribution
(EDO),
optical
microscope
(OMA),
secretary
bird
(SBOA),
Parrot
Optimizer
(PO),
Newton-Raphson-based
(NRBO),
crested
porcupine
(CPO),
differentiated
creative
(DCS),
propagation
(PSA),
one-to-one
(OOBO),
triangulation
topology
aggregation
(TTAO),
studied
clarify
effectiveness
models.
addition,
collaborate
functions,
namely
Lambert
W-Function
Newton-Raphson
Method,
aid
solving
I-V
curve
equations
more
accurately,
thereby
improving
Those
assessed
using
four
well-known
solar
cells
modules
compared
each
metrics,
best
fitness,
average
worst
standard
deviation
(SD),
Friedman
mean
rank,
speed;
multiple-comparison
test
compare
difference
between
ranks.
Results
comparison
show
that
SWO
efficient
effective
SDM,
DDM,
TDM
over
modules,
Method
equations.
study
reports
perform
poorly
when
applied
Heliyon,
Год журнала:
2024,
Номер
10(7), С. e28147 - e28147
Опубликована: Март 22, 2024
Deep
Convolutional
Neural
Networks
(DCNNs)
have
shown
remarkable
success
in
image
classification
tasks,
but
optimizing
their
hyperparameters
can
be
challenging
due
to
complex
structure.
This
paper
develops
the
Adaptive
Habitat
Biogeography-Based
Optimizer
(AHBBO)
for
tuning
of
DCNNs
tasks.
In
complicated
optimization
problems,
BBO
suffers
from
premature
convergence
and
insufficient
exploration.
this
regard,
an
adaptable
habitat
is
presented
as
a
solution
these
problems;
it
would
permit
variable
sizes
regulated
mutation.
Better
performance
greater
chance
finding
high-quality
solutions
across
wide
range
problem
domains
are
results
modification's
increased
exploration
population
diversity.
AHBBO
tested
on
53
benchmark
functions
demonstrates
its
effectiveness
improving
initial
stochastic
converging
faster
optimum.
Furthermore,
DCNN-AHBBO
compared
23
well-known
classifiers
nine
problems
shows
superior
reducing
error
rate
by
up
5.14%.
Our
proposed
algorithm
outperforms
13
87
out
95
evaluations,
providing
high-performance
reliable
DNNs
research
contributes
field
deep
learning
proposing
new
that
improve
efficiency
neural
networks
classification.
Applied Sciences,
Год журнала:
2025,
Номер
15(2), С. 608 - 608
Опубликована: Янв. 10, 2025
In
this
paper,
a
multi-strategy
enhanced
chimpanzee
optimization
algorithm
(MSEChOA)
acting
on
path
planning
for
delivery
vehicles
is
proposed
to
achieve
the
goal
of
shortening
global
lengths
unmanned
and
obtaining
safer
paths.
initialization
phase,
introduces
hybrid
good
point
set
chaos
strategy,
combining
advantages
both
enhance
randomness
homogeneity
initial
population.
After
that,
it
incorporates
benchmark
weight
strategy
Gaussian-modulated
cosine
factor
adaptively
adjust
parameters,
thus
balancing
local
search
capabilities
improving
efficiency.
end,
enhancer
(GEE)
further
capability
in
later
phases,
thereby
avoiding
optima.
Experiments
several
test
functions
show
that
MSEChOA
outperforms
traditional
ChOA
other
algorithms
accuracy
convergence
speed.
simulation
experiments,
shows
stronger
ability
computational
efficiency
simple
complex
environments,
proving
its
feasibility
superiority
field
planning.
