Mathematics,
Journal Year:
2024,
Volume and Issue:
12(24), P. 3921 - 3921
Published: Dec. 12, 2024
Accurate
sales
forecasting
is
essential
for
optimizing
resource
allocation,
managing
inventory,
and
maximizing
profit
in
competitive
markets.
Machine
learning
models
are
being
increasingly
used
to
develop
reliable
sales-forecasting
systems
due
their
advanced
capabilities
handling
complex
data
patterns.
This
study
introduces
a
novel
hybrid
approach
that
combines
the
artificial
bee
colony
(ABC)
fire
hawk
optimizer
(FHO)
algorithms,
specifically
designed
enhance
hyperparameter
optimization
machine
learning-based
models.
By
leveraging
strengths
of
these
two
metaheuristic
method
enhances
predictive
accuracy
robustness
models,
with
focus
on
hyperparameters
XGBoost
tasks.
Evaluations
across
three
distinct
datasets
demonstrated
model
consistently
outperformed
standalone
including
genetic
algorithm
(GA),
rabbits
(ARO),
white
shark
(WSO),
ABC
algorithm,
FHO,
latter
applied
first
time
optimization.
The
superior
performance
was
confirmed
through
RMSE,
MAPE,
statistical
tests,
marking
significant
advancement
providing
reliable,
effective
solution
refining
support
business
decision-making.
Mathematics,
Journal Year:
2024,
Volume and Issue:
12(22), P. 3502 - 3502
Published: Nov. 9, 2024
Rockburst
is
a
common
dynamic
geological
disaster
in
underground
mining
and
tunneling
engineering,
characterized
by
randomness,
abruptness,
impact.
Short-term
evaluation
of
rockburst
potential
plays
an
outsize
role
ensuring
the
safety
workers,
equipment,
projects.
As
well
known,
microseismic
monitoring
serves
as
reliable
short-term
early-warning
technique
for
rockburst.
However,
large
amount
data
brings
many
challenges
to
traditional
manual
analysis,
such
timeliness
processing
accuracy
prediction.
To
this
end,
study
integrates
artificial
intelligence
with
monitoring.
On
basis
comprehensive
consideration
class
imbalance
multicollinearity,
innovative
modeling
framework
that
combines
local
outlier
factor-guided
synthetic
minority
oversampling
extremely
randomized
forest
C5.0
decision
trees
proposed
potential.
determine
optimal
hyperparameters,
whale
optimization
algorithm
embedded.
prove
efficacy
model,
total
93
cases
are
collected
from
various
engineering
The
results
show
approach
achieves
90.91%
macro
F1-score
0.9141.
Additionally,
F1-scores
on
low-intensity
high-intensity
0.9600
0.9474,
respectively.
Finally,
advantages
further
validated
through
extended
comparative
analysis.
insights
derived
research
provide
reference
data-based
prediction
when
faced
multicollinearity.
Frontiers in Earth Science,
Journal Year:
2025,
Volume and Issue:
12
Published: Jan. 20, 2025
The
occurrence
of
class-imbalanced
datasets
is
a
frequent
observation
in
natural
science
research,
emphasizing
the
paramount
importance
effectively
harnessing
them
to
construct
highly
accurate
models
for
rockburst
prediction.
Initially,
genuine
incidents
within
burial
depth
500
m
were
sourced
from
literature,
revealing
small
dataset
imbalance
issue.
Utilizing
various
mainstream
oversampling
techniques,
was
expanded
generate
six
new
datasets,
subsequently
subjected
12
classifiers
across
84
classification
processes.
model
incorporating
highest-scoring
original
and
top
two
dataset,
yielded
high-performance
model.
Findings
indicate
that
KMeansSMOTE
technique
exhibits
most
substantial
enhancement
combined
classifiers,
whereas
individual
favor
ET+SVMSMOTE
RF+SMOTENC.
Following
multiple
rounds
hyper
parameter
adjustment
via
random
cross-validation,
combination
attained
highest
accuracy
rate
93.75%,
surpassing
Moreover,
SVMSMOTE
technique,
augmenting
samples
with
fewer
categories,
demonstrated
notable
benefits
mitigating
overfitting,
enhancing
generalization,
improving
Recall
F1
score
RF
classifiers.
Validated
its
high
generalization
performance,
accuracy,
reliability.
This
process
also
provides
an
efficient
framework
development.
