Water,
Год журнала:
2024,
Номер
16(24), С. 3656 - 3656
Опубликована: Дек. 18, 2024
In
recent
years,
frequent
floods
caused
by
heavy
rainfall
and
persistent
precipitation
have
greatly
affected
changes
in
groundwater
levels.
This
has
not
only
huge
economic
losses
human
casualties,
but
also
had
a
significant
impact
on
the
ecological
environment.
The
aim
of
this
study
is
to
explore
effectiveness
new
method
based
Long
Short-Term
Memory
networks
(LSTM)
its
optimization
model
level
prediction
compared
with
traditional
method,
evaluate
accuracy
different
models,
identify
main
factors
affecting
level.
Taking
Chaoyang
City
Liaoning
Province
as
an
example,
four
assessment
indicators,
R2,
MAE,
RMSE,
MAPE,
were
used.
results
show
that
optimized
LSTM
outperforms
both
underlying
all
metrics,
GWO-LSTM
performing
best.
It
was
found
high
water-table
anomalies
are
mainly
or
storms.
Changes
water
table
can
negatively
affect
environment
such
vegetation
growth,
soil
salinization,
geological
hazards.
accurate
levels
scientific
importance
for
development
sustainable
cities
communities,
well
good
health
well-being
beings.
Frontiers in Artificial Intelligence,
Год журнала:
2025,
Номер
7
Опубликована: Янв. 7, 2025
This
systematic
review
provides
a
state-of-art
of
Artificial
Intelligence
(AI)
models
such
as
Machine
Learning
(ML)
and
Deep
(DL)
development
its
applications
in
Mexico
diverse
fields.
These
are
recognized
powerful
tools
many
fields
due
to
their
capability
carry
out
several
tasks
forecasting,
image
classification,
recognition,
natural
language
processing,
machine
translation,
etc.
article
aimed
provide
comprehensive
information
on
the
algorithms
applied
Mexico.
A
total
120
original
research
papers
were
included
details
trends
publication,
spatial
location,
institutions,
publishing
issues,
subject
areas,
applied,
performance
metrics
discussed.
Furthermore,
future
directions
opportunities
presented.
15
areas
identified,
where
Social
Sciences
Medicine
main
application
areas.
It
observed
that
Neural
Networks
(ANN)
preferred,
probably
learn
model
non-linear
complex
relationships
addition
other
popular
Random
Forest
(RF)
Support
Vector
Machines
(SVM).
identified
selection
rely
study
objective
data
patterns.
Regarding
accuracy
recall
most
employed.
paper
could
assist
readers
understanding
techniques
used
area
field
country.
Moreover,
significant
knowledge
implementation
national
AI
strategy,
according
country
needs.
Water,
Год журнала:
2024,
Номер
16(21), С. 3125 - 3125
Опубликована: Ноя. 1, 2024
Flood
disasters
often
result
in
significant
losses
of
life
and
property,
making
them
among
the
most
devastating
natural
hazards.
Therefore,
reliable
accurate
water
level
forecasting
is
critically
important.
Rainfall-runoff
modeling,
which
a
complex
nonlinear
time
series
process,
plays
key
role
this
endeavor.
Numerous
studies
have
demonstrated
that
data-driven
methods,
particularly
deep
learning
approaches
such
as
convolutional
neural
networks
(CNNs),
long
short-term
memory
(LSTM)
networks,
transformers,
shown
promising
performance
prediction
tasks.
This
study
introduces
Conformer,
novel
architecture
integrates
strengths
CNNs
transformers
for
rainfall-runoff
modeling.
The
framework
uses
self-attention
mechanisms
combined
with
computations
to
extract
essential
features—such
levels,
precipitation,
meteorological
data—from
multiple
stations,
are
then
aggregated
predict
subsequent
series.
utilized
data
spanning
from
1
April
2006
25
July
2021,
totaling
5595
days
(134,280
h),
were
divided
into
training,
validation,
test
sets
an
8:1:1
ratio
train
model,
adjust
parameters,
evaluate
performance,
respectively.
effectiveness
feasibility
proposed
model
evaluated
Lanyang
River
Basin,
focus
on
predicting
7-day-ahead
levels.
results
obtained
ablation
experiments
indicate
significantly
enhance
ability
capture
local
relationships
between
levels
other
parameters.
Additionally,
performing
convolution
after
executing
operations
yields
even
better
results.
Compared
models
simulations,
Conformer
markedly
outperforms
CNN,
LSTM,
traditional
transformer
terms
coefficient
determination
(R2)
Nash–Sutcliffe
efficiency
(NSE)
indicators.
These
findings
highlight
potential
replace
commonly
used
methods
field
hydrology.
Journal of Soft Computing Paradigm,
Год журнала:
2024,
Номер
6(1), С. 55 - 69
Опубликована: Март 1, 2024
This
research
focuses
on
predicting
water
sources
in
various
areas
by
analyzing
historical
data
groundwater
levels,
rainfall,
and
borewells.
The
study
explores
the
relationships
between
levels
environmental
factors,
emphasizing
influence
of
rainfall
aquifer
recharge.
Borewell
data,
including
depth
quality,
is
incorporated
to
identify
potential
sources.
involves
cleaning,
exploratory
analysis,
machine
learning
predict
based
diverse
features
such
as
patterns
geographical
characteristics.
Spatial
analysis
using
GIS
tools
visualizes
distribution
rainfall.
model's
performance
evaluated,
considering
metrics
local
hydrogeological
conditions,
with
an
emphasis
integrating
borewell
data.
Continuous
monitoring
updates
ensure
ongoing
relevance.
integrated
approach
aims
provide
insights
for
sustainable
resource
management,
assisting
decision-makers
planning
areas.
