Sustainability,
Год журнала:
2024,
Номер
16(23), С. 10634 - 10634
Опубликована: Дек. 4, 2024
Accurately
predicting
the
state
of
surface
water
quality
is
crucial
for
ensuring
sustainable
use
resources
and
environmental
protection.
This
often
requires
a
focus
on
range
factors
affecting
quality,
such
as
physical
chemical
parameters.
Tree
models,
with
their
flexible
tree-like
structure
strong
capability
partitioning
selecting
influential
features,
offer
clear
decision-making
rules,
making
them
suitable
this
task.
However,
an
individual
decision
tree
model
has
limitations
cannot
fully
capture
complex
relationships
between
all
influencing
parameters
quality.
Therefore,
study
proposes
method
combining
ensemble
models
voting
algorithms
to
predict
classification.
was
conducted
using
five
monitoring
sites
in
Qingdao,
representing
portion
many
municipal
environment
stations
China,
employing
single-factor
determination
stringent
standards.
The
soft
algorithm
achieved
highest
accuracy
99.91%,
addressed
imbalance
original
categories,
reaching
Matthews
Correlation
Coefficient
(MCC)
99.88%.
In
contrast,
conventional
machine
learning
algorithms,
logistic
regression
K-nearest
neighbors,
lower
accuracies
75.90%
91.33%,
respectively.
Additionally,
model’s
supervision
misclassified
data
demonstrated
its
good
rules.
trained
also
transferred
directly
at
13
Beijing,
where
it
performed
robustly,
achieving
hard
97.73%
MCC
96.81%.
countries’
systems,
different
qualities
correspond
uses,
magnitude
related
categories;
critical
can
even
determine
category.
are
highly
capable
handling
nonlinear
important
allowing
identify
exploit
interactions
parameters,
which
especially
when
multiple
together
there
significant
motivation
develop
model-based
prediction
models.
Journal of Materials Chemistry A,
Год журнала:
2024,
Номер
12(32), С. 20717 - 20782
Опубликована: Янв. 1, 2024
Evaluating
the
advantages
and
limitations
of
applying
machine
learning
for
prediction
optimization
in
porous
media,
with
applications
energy,
environment,
subsurface
studies.
Environmental Earth Sciences,
Год журнала:
2023,
Номер
82(17)
Опубликована: Авг. 9, 2023
Abstract
Groundwater
quality
(GWQ)
monitoring
is
one
of
the
best
environmental
objectives
due
to
recent
droughts
and
urban
rural
development.
Therefore,
this
study
aimed
map
GWQ
in
central
plateau
Iran
by
validating
machine
learning
algorithms
(MLAs)
using
game
theory
(GT).
On
basis,
chemical
parameters
related
water
quality,
including
K
+
,
Na
Mg
2+
Ca
SO
4
2−
Cl
−
HCO
3
pH,
TDS,
EC,
were
interpolated
at
39
sampling
sites.
Then,
random
forest
(RF),
support
vector
(SVM),
Naive
Bayes,
K-nearest
neighbors
(KNN)
used
Python
programming
language,
was
plotted
concerning
GWQ.
Borda
scoring
validate
MLAs,
sample
points
prioritized.
Based
on
results,
among
ML
algorithms,
RF
algorithm
with
error
statistics
MAE
=
0.261,
MSE
0.111,
RMSE
0.333,
AUC
0.930
selected
as
most
optimal
algorithm.
created
algorithm,
42.71%
studied
area
poor
condition.
The
proportion
region
classes
moderate
high
18.93%
38.36%,
respectively.
results
prioritization
sites
GT
showed
a
great
similarity
between
model.
In
addition,
analysis
condition
critical
non-critical
based
that
aspects,
carbonate
balance,
salinity
general,
it
can
be
said
simultaneous
use
MLA
provides
good
basis
for
constructing
Iran.
Environmental Science & Technology Letters,
Год журнала:
2023,
Номер
10(9), С. 804 - 809
Опубликована: Авг. 29, 2023
Reactive
chlorine
species
(RCS),
such
as
(HOCl/OCl–),
dioxide
(ClO2),
atom
(Cl•),
and
dichlorine
radical
(Cl2•–),
play
a
crucial
role
in
oxidation
disinfection
worldwide.
In
this
study,
we
developed
machine
learning
(ML)-based
quantitative
structure–activity
relationship
(QSAR)
models
to
predict
the
rate
constants
of
RCS
toward
organic
compounds
by
using
quantum
chemical
descriptors
(QDs)
Morgan
fingerprints
(MFs)
input
features
along
with
three
tree-based
ML
algorithms.
The
ML-based
(RMSEtest
=
0.528–1.131)
outperform
multiple
linear
regression-based
0.772–4.837).
Moreover,
QSAR
combining
QDs
MFs
0.528–0.948)
show
better
prediction
performance
than
that
0.616–1.875)
or
alone
0.636–1.439)
for
all
four
RCS.
SHapely
Additive
exPlanation
(SHAP)
analysis
reveals
energy
highest
occupied
molecular
orbital
(EHOMO),
charge,
−O––NH2
−CO
are
most
important
affecting
This
study
demonstrates
combination
achieves
much
model
RCS,
which
can
be
extrapolated
other
oxidants
water
treatment.
