Hyperparameter
tuning
is
a
crucial
step
in
the
process
of
building
accurate
machine
learning
models.
Finding
optimal
combination
hyperparameters
can
be
challenging,
especially
complex
models
with
large
hyperparameter
spaces.
Genetic
algorithms
(GAs)
have
become
popular
approach
to
address
this
challenge
by
efficiently
exploring
space
and
selecting
best
combination.
In
paper,
different
are
used
along
genetic
search
for
prediction
multi-diseases.
The
purpose
using
algorithm
optimize
hyper-parameters.
Based
upon
evaluations
we
come
know
which
performed
well
after
hyper-parameter
optimization
meta
heuristic
optimization.
Diagnostics,
Год журнала:
2022,
Номер
12(11), С. 2718 - 2718
Опубликована: Ноя. 7, 2022
In
the
last
few
years,
artificial
intelligence
has
shown
a
lot
of
promise
in
medical
domain
for
diagnosis
and
classification
human
infections.
Several
computerized
techniques
based
on
(AI)
have
been
introduced
literature
gastrointestinal
(GIT)
diseases
such
as
ulcer,
bleeding,
polyp,
others.
Manual
these
infections
is
time
consuming,
expensive,
always
requires
an
expert.
As
result,
methods
that
can
assist
doctors
second
opinion
clinics
are
widely
required.
The
key
challenges
technique
accurate
infected
region
segmentation
because
each
change
shape
location.
Moreover,
inaccurate
affects
feature
extraction
later
impacts
accuracy.
this
paper,
we
proposed
automated
framework
GIT
disease
deep
saliency
maps
Bayesian
optimal
learning
selection.
made
up
steps,
from
preprocessing
to
classification.
Original
images
improved
step
by
employing
contrast
enhancement
technique.
following
step,
map
segmenting
regions.
segmented
regions
then
used
train
pre-trained
fine-tuned
model
called
MobileNet-V2
using
transfer
learning.
model's
hyperparameters
were
initialized
optimization
(BO).
average
pooling
layer
extract
features.
However,
several
redundant
features
discovered
during
analysis
phase
must
be
removed.
hybrid
whale
algorithm
selecting
best
Finally,
selected
classified
extreme
machine
classifier.
experiment
was
carried
out
three
datasets:
Kvasir
1,
2,
CUI
Wah.
achieved
accuracy
98.20,
98.02,
99.61%
datasets,
respectively.
When
compared
other
methods,
shows
improvement
