The
time
series
analysis
of
clinical
data
in
health
sciences
is
considered
as
an
important
study
the
field
computer
science.
more
accurate
future
predicted,
vital
measures
for
human
can
be
taken
and
costs
sector
minimized.
main
purpose
algorithms
to
predict
a
certain
interval
based
on
past
data.
However,
instead
using
pure
versions
algorithms,
it
necessary
combine
these
with
new
approaches
improve
their
performance
well-known
increase
prediction
accuracy.
Therefore,
this
study,
twenty-six
different
have
been
used
forecasting
two
datasets.
top
five
best
results
chosen
from
algorithms.
Then,
ensemble
learning
model
proposed
by
applying
mean,
median
voting
techniques
datasets
are
COVID-19
transmission
dataset
Turkey
hospital
Emergency
Department
arrival
Iowa.
As
result
experimental
studies,
96.10%
accuracy
value
92.88%
ED
achieved,
respectively.
It
has
observed
that
achieves
better
compared
LSTM
deep
architecture.
In
conclusion,
shown
problems
Natural Language Processing Journal,
Год журнала:
2024,
Номер
7, С. 100067 - 100067
Опубликована: Март 22, 2024
The
infectious
diseases,
such
as
COVID-19
pandemic,
has
led
to
a
surge
of
information
on
the
internet,
including
misinformation,
necessitating
fact-checking
tools.
However,
diseases
related
claims
pose
challenges
due
informal
versus
formal
evidence
and
presence
multiple
aspects
in
claim.
To
address
these
issues,
we
propose
soft
prompt-based
ensemble
learning
framework
for
fact
checking.
understand
complex
assertions
social
media
texts,
explore
various
prompt
structures
take
advantage
T5
language
model,
together.
Soft
prompts
offer
flexibility
better
generalization
compared
hard
prompts.
model
captures
linguistic
cues
contextual
COVID-19-related
data,
thus
enhances
new
claims.
Experimental
results
demonstrate
that
improves
accuracy
provides
promising
approach
combat
misinformation
during
pandemic.
In
addition,
method
also
shows
great
zero-shot
capability
can
be
applied
checking
problems.
Future Internet,
Год журнала:
2024,
Номер
16(8), С. 286 - 286
Опубликована: Авг. 8, 2024
The
proliferation
of
fake
news
poses
a
significant
challenge
in
today’s
information
landscape,
spanning
diverse
domains
and
topics
undermining
traditional
detection
methods
confined
to
specific
domains.
In
response,
there
is
growing
interest
strategies
for
detecting
cross-domain
misinformation.
However,
machine
learning
(ML)
approaches
often
struggle
with
the
nuanced
contextual
understanding
required
accurate
classification.
To
address
these
challenges,
we
propose
novel
contextualized
prompt-based
zero-shot
approach
utilizing
pre-trained
Generative
Pre-trained
Transformer
(GPT)
model
(FND).
contrast
conventional
fine-tuning
reliant
on
extensive
labeled
datasets,
our
places
particular
emphasis
refining
prompt
integration
classification
logic
within
model’s
framework.
This
refinement
enhances
ability
accurately
classify
across
Additionally,
adaptability
allows
customization
tasks
by
modifying
placeholders.
Our
research
significantly
advances
demonstrating
efficacy
methodologies
text
classification,
particularly
scenarios
limited
training
data.
Through
experimentation,
illustrate
that
method
effectively
captures
domain-specific
features
generalizes
well
other
domains,
surpassing
existing
models
terms
performance.
These
findings
contribute
ongoing
efforts
combat
dissemination,
environments
severely
data,
such
as
online
platforms.
