Predicting preterm birth using machine learning methods
Scientific Reports,
Год журнала:
2025,
Номер
15(1)
Опубликована: Фев. 16, 2025
Preterm
birth
is
a
significant
public
health
concern,
given
its
correlation
with
neonatal
mortality
and
morbidity.
The
aetiology
of
preterm
complex
multifactorial.
objective
this
study
was
to
develop
compare
machine
learning
models
for
predicting
the
risk
birth.
Data
were
collected
from
50
patients
in
maternity
ward,
an
analysis
performed
based
on
timing
delivery
(preterm
vs.
term).
applicability
XGBoost,
CatBoost,
logistic
regression,
support
vector
machines
(SVM),
decision
trees
evaluated
through
training.
linear
SVM
boosted
parameters
demonstrated
highest
performance,
achieving
accuracy
82%,
precision
83%,
recall
86%,
F1-score
84%.
regression
model,
also
boosted,
comparable
performance
SVM,
similar
(80%),
(82%),
(82%).
other
models,
including
more
algorithms,
inferior,
which
likely
attributable
limited
dataset
number
involved.
In
particular,
most
notably
can
be
effectively
employed
assess
findings
indicate
that
model
exhibits
greatest
efficacy
among
tested
models.
Язык: Английский
Developing a logistic regression model to predict spontaneous preterm birth from maternal socio-demographic and obstetric history at initial pregnancy registration
BMC Pregnancy and Childbirth,
Год журнала:
2024,
Номер
24(1)
Опубликована: Окт. 21, 2024
Abstract
Background
Current
predictive
machine
learning
techniques
for
spontaneous
preterm
birth
heavily
rely
on
a
history
of
previous
and/or
costly
such
as
fetal
fibronectin
and
ultrasound
measurement
cervical
length
to
the
disadvantage
those
considered
at
low
risk
who
have
no
access
more
expensive
screening
tools.
Aims
objectives
We
aimed
develop
model
delivery
<
37
weeks
using
socio-demographic
clinical
data
readily
available
booking
-an
approach
which
could
be
suitable
all
women
regardless
their
obstetric
history.
Methods
developed
logistic
regression
seven
feature
variables
derived
from
maternal
(
n
=
917)
matched
full-term
100)
cohort
in
2018
2020
tertiary
unit
UK.
A
three-fold
cross-validation
technique
was
applied
with
subsets
training
testing
Python®
(version
3.8)
most
factors.
The
performance
then
compared
previously
published
algorithms.
Results
retrospective
showed
good
accuracy
an
AUC
0.76
(95%
CI:
0.71–0.83)
birth,
sensitivity
specificity
0.71
0.66–0.76)
0.78
0.63–0.88)
respectively
based
variables:
age,
BMI,
ethnicity,
smoking,
gestational
type,
substance
misuse
parity/obstetric
Conclusion
Pending
further
validation,
our
observations
suggest
that
key
demographic
features,
incorporated
into
traditional
mathematical
model,
promising
utility
pregnant
region
without
need
fibronectin.
Язык: Английский
Application of ITransformers to Predicting Preterm Birth Rate. Comparison with the ARIMA Model
Metody Ilościowe w Badaniach Ekonomicznych,
Год журнала:
2024,
Номер
25(3), С. 124 - 133
Опубликована: Сен. 30, 2024
Язык: Английский
Data Flow-Based Strategies to Improve the Interpretation and Understanding of Machine Learning Models
Bioengineering,
Год журнала:
2024,
Номер
11(12), С. 1189 - 1189
Опубликована: Ноя. 25, 2024
Data
flow-based
strategies
that
seek
to
improve
the
understanding
of
A.I.-based
results
are
examined
here
by
carefully
curating
and
monitoring
flow
data
into,
for
example,
artificial
neural
networks
random
forest
supervised
models.
While
these
models
possess
structures
related
fitting
procedures
highly
complex,
careful
restriction
being
utilized
can
provide
insight
into
how
they
interpret
associated
variables
sets
affected
differing
levels
variation
in
data.
The
goal
is
improving
our
modeling-based
their
stability
across
different
sources.
Some
guidelines
suggested
such
first-stage
adjustments
issues.
Язык: Английский