Forests,
Journal Year:
2025,
Volume and Issue:
16(2), P. 329 - 329
Published: Feb. 13, 2025
Fire
risk
prediction
is
of
great
importance
for
fire
prevention.
maps
are
an
effective
tool
to
quantify
regional
risk.
Most
existing
studies
on
forest
mainly
use
a
single
machine
learning
model,
but
different
models
have
varying
degrees
feature
extraction
in
the
same
spatial
environment,
leading
inconsistencies
accuracy.
To
address
this
issue,
study
proposes
novel
integrated
framework
that
systematically
evaluates
multiple
and
combines
their
outputs
through
weighted
ensemble
approach,
thereby
enhancing
robustness.
During
selection
stage,
factors
including
socio-economic,
climate,
terrain,
remote
sensing
data,
human
were
considered.
Unlike
previous
eight
evaluated
compared
using
performance
metrics.
Three
based
Mean
Squared
Error
(MSE)
values,
cross-validation
results
showed
improvement
model
performance.
The
achieved
accuracy
0.8602,
area
under
curve
(AUC)
0.772,
superior
sensitivity
(0.9234),
outperforming
individual
models.
Finally,
was
applied
generate
map.
Compared
with
prior
studies,
multi-model
approach
not
only
improves
predictive
also
provides
scalable
adaptable
mapping,
valuable
insights
future
sustainability
issues.
Remote Sensing,
Journal Year:
2025,
Volume and Issue:
17(3), P. 547 - 547
Published: Feb. 6, 2025
Soil,
a
crucial
natural
resource
and
the
cornerstone
of
agriculture,
profoundly
impacts
crop
growth,
quality,
yield.
However,
soil
degradation
affects
over
one-third
global
land,
with
topsoil
loss
emerging
as
significant
form
this
degradation,
posing
grave
threat
to
agricultural
sustainability
socio-economic
development.
Therefore,
accurate
monitoring
topsoil-loss
distribution
is
essential
for
formulating
effective
protection
management
strategies.
Traditional
survey
methods
are
limited
by
time-consuming
labor-intensive
processes,
high
costs,
complex
data
processing.
These
limitations
make
it
particularly
challenging
meet
demands
large-scale
research
efficient
information
imperative
develop
more
extraction
method.
This
study
focuses
on
Heshan
Farm
in
Heilongjiang
Province,
China,
subject
utilizes
remote
sensing
technology
machine
learning
methods.
It
introduces
multi-source
data,
including
Sentinel-2
satellite
imagery
Digital
Elevation
Model
(DEM)
design
four
schemes.
(1)
spectral
feature
extraction;
(2)
+
topographic
(3)
index
(4)
extraction.
Models
identification
based
Random
Forest
(RF)
Support
Vector
Machine
(SVM)
algorithms
developed,
Particle
Swarm
Optimization
(PSO)
algorithm
introduced
optimize
models.
The
performance
models
evaluated
using
overall
accuracy
Kappa
coefficient
indicators.
results
show
that
Scheme
4,
which
integrates
features,
various
indices,
performs
best
effects.
RF
model
demonstrates
higher
classification
than
SVM
model.
optimized
PSO-RF
PSO-SVM
improvements
accuracy,
especially
model,
an
0.97
0.94.
4
improves
OA
34.72%
38.81%
compared
1.
Topsoil
has
negative
impact
severely
restricting
normal
growth
development
crops.
provides
technical
means
black-soil
regions
offers
scientific
basis
ecological
strategies,
thereby
promoting
sustainable
resources.
Energies,
Journal Year:
2025,
Volume and Issue:
18(4), P. 842 - 842
Published: Feb. 11, 2025
Accurate
oil
and
gas
production
forecasting
is
essential
for
optimizing
field
development
operational
efficiency.
Steady-state
capacity
prediction
models
based
on
machine
learning
techniques,
such
as
Linear
Regression,
Support
Vector
Machines,
Random
Forest,
Extreme
Gradient
Boosting,
effectively
address
complex
nonlinear
relationships
through
feature
selection,
hyperparameter
tuning,
hybrid
integration,
achieving
high
accuracy
reliability.
These
maintain
relative
errors
within
acceptable
limits,
offering
robust
support
reservoir
management.
Recent
advancements
in
spatiotemporal
modeling,
Physics-Informed
Neural
Networks
(PINNs),
agent-based
modeling
have
further
enhanced
transient
forecasting.
Spatiotemporal
capture
temporal
dependencies
spatial
correlations,
while
PINN
integrates
physical
laws
into
neural
networks,
improving
interpretability
robustness,
particularly
sparse
or
noisy
data.
Agent-based
complements
these
techniques
by
combining
measured
data
with
numerical
simulations
to
deliver
real-time,
high-precision
predictions
of
dynamics.
Despite
challenges
computational
scalability,
sensitivity,
generalization
across
diverse
reservoirs,
future
developments,
including
multi-source
lightweight
architectures,
real-time
predictive
capabilities,
can
improve
forecasting,
addressing
the
complexities
supporting
sustainable
resource
management
global
energy
security.
Micromachines,
Journal Year:
2025,
Volume and Issue:
16(5), P. 507 - 507
Published: April 26, 2025
This
study
proposes
a
novel
parallel
denoising
and
temperature
compensation
fusion
algorithm
for
MEMS
ring
gyroscopes.
First,
the
particle
swarm
optimization
(PSO)
is
used
to
optimize
time-varying
filter-based
empirical
mode
decomposition
(TVFEMD),
obtaining
optimal
parameters.
Then,
TVFEMD
decomposes
gyroscope
output
signal
into
series
of
product
function
(PF)
signals
residual
signal.
Next,
sample
entropy
(SE)
employed
classify
decomposed
three
categories:
noise
segment,
mixed
feature
segment.
According
model
structure,
segment
directly
discarded.
Meanwhile,
time–frequency
peak
filtering
(TFPF)
applied
denoise
while
undergoes
compensation.
For
compensation,
football
team
training
(FTTA)
parameters
long
short-term
memory
(LSTM)
neural
network,
forming
FTTA-LSTM
architecture.
Both
simulations
experimental
results
validate
effectiveness
proposed
algorithm.
After
processing
using
PSO-TVFEMD-SE-TFPF
drift
model,
angular
random
walk
(ARW)
reduced
0.02°/√h,
bias
instability
(BI)
decreases
2.23°/h.
Compared
original
signal,
ARW
BI
are
by
99.43%
97.69%,
respectively.
The
fusion-based
method
significantly
enhances
stability
performance
gyroscope.
Scientific Reports,
Journal Year:
2025,
Volume and Issue:
15(1)
Published: May 21, 2025
Abstract
Power
Line
Communication
(PLC)
facilitates
the
usage
of
power
cables
to
transmit
data.
The
issue
is
that
sending
data
unavailable
nodes
time-consuming.
Machine
Learning
has
solved
this
by
predicting
a
node
having
optimum
readings.
more
machine
learning
models
learn,
accurate
they
become,
as
model
becomes
always
updated
with
node’s
continuous
availability
status,
so
self-training
algorithms
have
been
used.
A
dataset
2000
instances
500-node
implemented
PLC
network
collected.
These
consist
CINR(Carrier-to-Interference
plus
Noise
Ratio),
SNR(Signal-to-Noise
and
RSSI(Received
Signal
Strength
Indicator)
features
for
label,
which
UP/Down.
set
split
into
85%
training
15%
testing
set.
are
unlabeled.
Self-training
classifier
used
allow
Light
Gradient
Boosting
(LGBM)
Support
Vector
(linear
non-linear
kernel)
behave
in
manner
well
label
propagation
spreading
algorithms.
Supervised
(Random
Forest
logistic
regression)
trained
on
compare
results.
best
Label
Spreading,
resulted
accuracy
equals
94.67%,
f1-score
0.947,
precision
0.946,
recall
0.947
time
0.018
sec.
memory
consumption
0.99
MB.
