Research Square (Research Square),
Journal Year:
2023,
Volume and Issue:
unknown
Published: July 19, 2023
Abstract
In
this
study,
machine
learning
classifiers
are
integrated
with
the
geostatistical
analyses.
The
data
extracted
from
surface
maps
derived
ordinary
kriging
were
passed
onto
ML
algorithms,
resulting
in
prediction
accuracies
of
95%
(Gradient
Boosting
Classifier)
for
classification
and
91%
(Random
Forest
Regressor)
Regression.
Kmeans
clustering
model
provided
better
results
analysis
based
on
Silhouette,
Calinski-Harabasz,
Davies-Bouldin
metrics.
However,
there
was
certain
overfitting
prediction,
probably
due
to
limited
available
analysis.
addition,
interpolation
methods
might
have
affected
performance
by
producing
underfitting
results.
It
is
report
that
Gradient
classifier
mode
yielded
relatively
high
predicting
groundwater
quality
when
three
classes
used.
Random
Regressor
regression
returned
features
multiple
used
study.
This
work
reports
algorithms
can
predict
minimal
expense
expertise.
Frontiers in Water,
Journal Year:
2024,
Volume and Issue:
6
Published: April 18, 2024
Groundwater
models
often
require
transmissivity
(
T
)
fields
as
an
input.
These
are
commonly
generated
by
performing
univariate
interpolation
of
the
data.
This
data
is
derived
from
pumping
tests
and
generally
limited
due
to
large
costs
logistical
requirements.
Hence
using
this
may
not
be
representative
for
a
whole
study
region.
(using
kriging,
IDW
etc.)
Hence,
presents
novel
cokriging
based
methodology
generate
credible
fields.
Cokriging
-
multivariate
geostatistical
method
permits
incorporation
additional
correlated
auxiliary
variables
generation
enhanced
Here
abundantly
available
litholog
saturated
thickness
has
been
used
secondary
(auxiliary)
given
its
correlation
with
primary
Additionally,
proposed
addresses
two
operational
problems
traditional
procedure.
The
first
problem
poor
estimation
variogram
cross-variogram
parameters
sparse
second
determination
relative
contributions
variable
in
process.
have
resolved
proposing
set
non-bias
conditions,
linking
interpolator
head
inverse
solution
these
parameters.
applied
Bist
doab
region
Punjab
(India).
base
line
studies
performed
elucidate
superiority
over
kriging
terms
reproducibility.
Research Square (Research Square),
Journal Year:
2023,
Volume and Issue:
unknown
Published: July 19, 2023
Abstract
In
this
study,
machine
learning
classifiers
are
integrated
with
the
geostatistical
analyses.
The
data
extracted
from
surface
maps
derived
ordinary
kriging
were
passed
onto
ML
algorithms,
resulting
in
prediction
accuracies
of
95%
(Gradient
Boosting
Classifier)
for
classification
and
91%
(Random
Forest
Regressor)
Regression.
Kmeans
clustering
model
provided
better
results
analysis
based
on
Silhouette,
Calinski-Harabasz,
Davies-Bouldin
metrics.
However,
there
was
certain
overfitting
prediction,
probably
due
to
limited
available
analysis.
addition,
interpolation
methods
might
have
affected
performance
by
producing
underfitting
results.
It
is
report
that
Gradient
classifier
mode
yielded
relatively
high
predicting
groundwater
quality
when
three
classes
used.
Random
Regressor
regression
returned
features
multiple
used
study.
This
work
reports
algorithms
can
predict
minimal
expense
expertise.
