Applied Sciences,
Год журнала:
2023,
Номер
13(4), С. 2353 - 2353
Опубликована: Фев. 11, 2023
Data-driven
analysis
and
characterization
of
molecular
phenotypes
comprises
an
efficient
way
to
decipher
complex
disease
mechanisms.
Using
emerging
next
generation
sequencing
technologies,
important
disease-relevant
outcomes
are
extracted,
offering
the
potential
for
precision
diagnosis
therapeutics
in
progressive
disorders.
Single-cell
RNA
(scRNA-seq)
allows
inherent
heterogeneity
between
individual
cellular
environments
be
exploited
provides
one
most
promising
platforms
quantifying
cell-to-cell
gene
expression
variability.
However,
high-dimensional
nature
scRNA-seq
data
poses
a
significant
challenge
downstream
analysis,
particularly
identifying
genes
that
dominant
across
cell
populations.
Feature
selection
is
crucial
step
reducing
dimensionality
facilitating
identification
relevant
biological
question.
Herein,
we
present
need
ensemble
feature
methodology
data,
specifically
context
Alzheimer’s
(AD).
We
combined
various
strategies
obtain
differentially
expressed
(DEGs)
AD
dataset,
providing
approach
identify
transcriptome
biomarkers
through
which
can
applied
other
diseases.
anticipate
techniques,
such
as
our
methodology,
will
dominate
options
especially
datasets
increase
volume
complexity,
leading
more
accurate
classification
features.
Heliyon,
Год журнала:
2024,
Номер
10(2), С. e24192 - e24192
Опубликована: Янв. 1, 2024
The
FOX
algorithm
is
a
recently
developed
metaheuristic
approach
inspired
by
the
behavior
of
foxes
in
their
natural
habitat.
While
exhibits
commendable
performance,
its
basic
version,
complex
problem
scenarios,
may
become
trapped
local
optima,
failing
to
identify
optimal
solution
due
weak
exploitation
capabilities.
This
research
addresses
high-dimensional
feature
selection
problem.
In
selection,
most
informative
features
are
retained
while
discarding
irrelevant
ones.
An
enhanced
version
proposed,
aiming
mitigate
drawbacks
selection.
improved
referred
as
S-shaped
Grey
Wolf
Optimizer-based
(FOX-GWO),
which
focuses
on
augmenting
search
capabilities
via
integration
GWO.
Additionally,
introduction
an
transfer
function
enables
population
explore
both
binary
options
throughout
process.
Through
series
experiments
18
datasets
with
varying
dimensions,
FOX-GWO
outperforms
83.33
%
for
average
accuracy,
61.11
reduced
dimensionality,
and
72.22
fitness
value
across
datasets.
Meaning
it
efficiently
explores
spaces.
These
findings
highlight
practical
potential
advance
data
analysis,
enhancing
model
prediction
accuracy.
Journal of Activity Sedentary and Sleep Behaviors,
Год журнала:
2024,
Номер
3(1)
Опубликована: Янв. 30, 2024
Abstract
The
nature
of
human
movement
and
non-movement
behaviors
is
complex
multifaceted,
making
their
study
complicated
challenging.
Thanks
to
the
availability
wearable
activity
monitors,
we
can
now
monitor
full
spectrum
physical
activity,
sedentary,
sleep
better
than
ever
before—whether
subjects
are
elite
athletes,
children,
adults,
or
individuals
with
pre-existing
medical
conditions.
increasing
volume
generated
data,
combined
inherent
complexities
behaviors,
necessitates
development
new
data
analysis
methods
for
research
behaviors.
characteristics
machine
learning
(ML)
methods,
including
ability
deal
make
them
suitable
such
thus
be
an
alternative
tool
this
nature.
ML
potentially
excellent
solving
many
traditional
problems
related
as
recognition,
posture
detection,
profile
analysis,
correlates
research.
However,
despite
potential,
has
not
yet
been
widely
utilized
analyzing
studying
these
In
review,
aim
introduce
experts
in
sedentary
behavior,
research—individuals
who
may
possess
limited
familiarity
ML—to
potential
applications
techniques
data.
We
begin
by
explaining
underlying
principles
modeling
pipeline,
highlighting
challenges
issues
that
need
considered
when
applying
ML.
then
present
types
ML:
supervised
unsupervised
learning,
a
few
algorithms
frequently
used
learning.
Finally,
highlight
three
areas
where
methodologies
have
already
behavior
research,
emphasizing
successes
challenges.
This
paper
serves
resource
offering
guidance
resources
facilitate
its
utilization.
IEEE Transactions on Fuzzy Systems,
Год журнала:
2024,
Номер
32(8), С. 4270 - 4284
Опубликована: Апрель 25, 2024
Multivariate
time
series
prediction
(MTSP)
stands
as
a
significant
and
challenging
frontier
in
the
data
science
domain,
garnering
considerable
interest
among
researchers.
Extreme
learning
machine
(ELM)
has
emerged
popular
algorithm
capable
of
effectively
addressing
MTSP
challenges.
However,
high-dimensional
nonlinear
nature
information
within
big
contexts
exposes
certain
limitations
ELM's
performance.
To
address
this
issue,
paper
proposes
hybrid
framework
based
on
fuzzy
C-means
(FCM)
clustering
coupled
with
feature
selection.
The
begins
possibility
distribution
(PD)-based
selection
designed
to
evaluate
quality
describe
uncertainty
via
multi-source
fusion.
Subsequently,
robust
FCM
is
developed,
optimizing
process
by
incorporating
differences
neighbor
samples
while
employing
multi-metric
strategy
determine
cluster
numbers.
Additionally,
an
enhanced
dual-kernel
ELM
(EDKELM)
network
established
enhance
capabilities.
resulting
excels
autonomously
discovering
intrinsic
featuremodel
connections,
exhibiting
superior
performance,
demonstrating
excellent
generalization
ability.
Experimental
results
using
real-world
datasets
showcase
competitiveness
proposed
over
existing
models
resolving
multivariate
Abstract
The
exponential
growth
of
digital
text
documents
presents
a
significant
challenge
for
classification
algorithms,
as
the
vast
number
words
in
each
document
can
hinder
their
efficiency.
Feature
selection
(FS)
is
crucial
technique
that
aims
to
eliminate
irrelevant
features
and
enhance
accuracy.
In
this
study,
we
propose
an
improved
version
discrete
laying
chicken
algorithm
(IDLCA)
utilizes
noun‐based
filtering
reduce
improve
performance.
Although
LCA
newly
proposed
algorithm,
it
has
not
been
systematically
applied
problems
before.
Our
enhanced
employs
different
operators
both
exploration
exploitation
find
better
solutions
mode.
To
evaluate
effectiveness
method,
compared
with
some
conventional
nature‐inspired
feature
methods
using
various
learning
models
such
decision
trees
(DT),
K‐nearest
neighbor
(KNN),
Naive
Bayes
(NB),
support
vector
machine
(SVM)
on
five
benchmark
datasets
three
evaluation
metrics.
experimental
results
demonstrate
comparison
existing
one.
code
available
at
https://github.com/m0javad/Improved-Discrete-Laying-Chicken-Algorithm
.
PLoS ONE,
Год журнала:
2024,
Номер
19(1), С. e0295579 - e0295579
Опубликована: Янв. 2, 2024
This
paper
proposes
a
feature
selection
method
based
on
hybrid
optimization
algorithm
that
combines
the
Golden
Jackal
Optimization
(GJO)
and
Grey
Wolf
Optimizer
(GWO).
The
primary
objective
of
this
is
to
create
an
effective
data
dimensionality
reduction
technique
for
eliminating
redundant,
irrelevant,
noisy
features
within
high-dimensional
datasets.
Drawing
inspiration
from
Chinese
idiom
“Chai
Lang
Hu
Bao,”
mechanisms,
cooperative
behaviors
observed
in
natural
animal
populations,
we
amalgamate
GWO
algorithm,
Lagrange
interpolation
method,
GJO
propose
multi-strategy
fusion
GJO-GWO
algorithm.
In
Case
1,
addressed
eight
complex
benchmark
functions.
2,
was
utilized
tackle
ten
problems.
Experimental
results
consistently
demonstrate
under
identical
experimental
conditions,
whether
solving
functions
or
addressing
problems,
exhibits
smaller
means,
lower
standard
deviations,
higher
classification
accuracy,
reduced
execution
times.
These
findings
affirm
superior
performance,
stability
