Current Bioinformatics,
Год журнала:
2022,
Номер
18(2), С. 109 - 142
Опубликована: Сен. 20, 2022
Background:
Moth-flame
optimization
will
meet
the
premature
and
stagnation
phenomenon
when
encountering
difficult
tasks.
Objective:
To
overcome
above
shortcomings,
this
paper
presented
a
quasi-reflection
moth-flame
algorithm
with
refraction
learning
called
QRMFO
to
strengthen
property
of
ordinary
MFO
apply
it
in
various
application
fields.
Method:
In
proposed
QRMFO,
quasi-reflection-based
increases
diversity
population
expands
search
space
on
iteration
jump
phase;
improves
accuracy
potential
optimal
solution.
Results:
Several
experiments
are
conducted
evaluate
superiority
paper;
first
all,
CEC2017
benchmark
suite
is
utilized
estimate
capability
dealing
standard
test
sets
compared
state-of-the-art
algorithms;
afterward,
adopted
deal
multilevel
thresholding
image
segmentation
problems
real
medical
diagnosis
case.
Conclusion:
Simulation
results
discussions
show
that
optimizer
superior
basic
other
advanced
methods
terms
convergence
rate
solution
accuracy.
Computational Intelligence and Neuroscience,
Год журнала:
2021,
Номер
2021(1)
Опубликована: Янв. 1, 2021
Cancer
can
be
considered
as
one
of
the
leading
causes
death
widely.
One
most
effective
tools
to
able
handle
cancer
diagnosis,
prognosis,
and
treatment
is
by
using
expression
profiling
technique
which
based
on
microarray
gene.
For
each
data
point
(sample),
gene
usually
receives
tens
thousands
genes.
As
a
result,
this
large‐scale,
high‐dimensional,
highly
redundant.
The
classification
profiles
(NP)‐Hard
problem.
Feature
(gene)
selection
methods
A
hybrid
approach
presented
in
paper,
several
machine
learning
techniques
were
used
model:
Pearson’s
correlation
coefficient
correlation‐based
feature
selector
reducer,
Decision
Tree
classifier
that
easy
interpret
does
not
require
parameter,
Grid
Search
CV
(cross‐validation)
optimize
maximum
depth
hyperparameter.
Seven
standard
datasets
are
evaluate
our
model.
To
identify
features
informative
relative
proposed
model,
various
performance
measurements
employed,
including
accuracy,
specificity,
sensitivity,
F
1‐score,
AUC.
suggested
strategy
greatly
decreases
number
genes
required
for
classification,
selects
features,
increases
according
results.
Current Bioinformatics,
Год журнала:
2022,
Номер
18(2), С. 109 - 142
Опубликована: Сен. 20, 2022
Background:
Moth-flame
optimization
will
meet
the
premature
and
stagnation
phenomenon
when
encountering
difficult
tasks.
Objective:
To
overcome
above
shortcomings,
this
paper
presented
a
quasi-reflection
moth-flame
algorithm
with
refraction
learning
called
QRMFO
to
strengthen
property
of
ordinary
MFO
apply
it
in
various
application
fields.
Method:
In
proposed
QRMFO,
quasi-reflection-based
increases
diversity
population
expands
search
space
on
iteration
jump
phase;
improves
accuracy
potential
optimal
solution.
Results:
Several
experiments
are
conducted
evaluate
superiority
paper;
first
all,
CEC2017
benchmark
suite
is
utilized
estimate
capability
dealing
standard
test
sets
compared
state-of-the-art
algorithms;
afterward,
adopted
deal
multilevel
thresholding
image
segmentation
problems
real
medical
diagnosis
case.
Conclusion:
Simulation
results
discussions
show
that
optimizer
superior
basic
other
advanced
methods
terms
convergence
rate
solution
accuracy.
