The
recent
emergence
of
time
series
contrastive
clustering
methods
can
be
broadly
categorized
into
two
classes.
first
class
uses
learning
to
learn
universal
representations
for
series.
Though
they
perform
well
in
various
downstream
tasks,
such
disregard
the
important
categorical
information
and
objective,
leading
unsuitable
tasks.
second
incorporates
objective.
potential
connections
structures
between
data
are
not
fully
explored
during
learning.
To
this
end,
we
propose
a
graph-augmented
framework
called
"Time
Series
Graph-augmented
Contrastive
Clustering
(TSGCC)
method."
We
observed
that
original
samples
should
similar
their
augmentations
other
same
cluster.
Hence,
used
weighted
$KNN$
graph
build
positive
negative
sample
pairs
Subsequently,
projected
instance
feature
space
with
dimensionality
number
clusters
learned
cluster-friendly
features
cluster
assignments
by
iteratively
optimizing
loss.
Experimental
results
demonstrate
TSGCC
outperforms
16
advanced
time-series
on
36
challenging
UCR
benchmarks,
achieving
best
12
datasets
highest
average
rank
(2.83)
RI
overall
methods.
ABSTRACT
Grid
clustering
is
a
classical
algorithm
with
the
advantage
of
lower
time
complexity,
which
suitable
for
analysis
large
geographic
data.
However,
it
sensitive
to
grid
division
parameter
M
and
density
threshold
R
,
accuracy
poor.
The
article
proposes
hybrid
HCA‐BGP
based
on
division.
first
uses
obtain
core
part
class
family,
then
division‐based
method
edge
family.
Through
experiments
simulated
datasets
real
datasets,
proved
have
better
results
than
existing
as
well
some
other
algorithms.
In
terms
accuracy,
compared
Clique,
F‐value
this
paper's
improved
by
20.3%
dataset
S1,
81.8%
R15,
7.6%
average
eight
datasets.
sensitivity
parameters
variance
clustered
reduced
89.3%
S1;
ARI
99.9%
Data8.
Compared
another
grid‐based
algorithm,
GDB,
also
demonstrates
significant
advantages.
PLoS ONE,
Год журнала:
2025,
Номер
20(5), С. e0318667 - e0318667
Опубликована: Май 12, 2025
The
drilling
process
is
a
crucial
component
in
the
production
of
panel
furniture
enterprises;
simultaneously,
it
also
most
complex
process.
And
enterprise’s
transition
to
intelligent
manufacturing
lacks
effective
optimization.
Therefore,
this
study
focuses
on
optimizing
furniture.
Initially,
an
analysis
cabinet
structures
was
conducted,
followed
by
data
collection
patterns.
Based
and
insights
from
hole
distribution
patterns,
novel
COING
(Coordinate
Information
Grid)
method
proposed.
Subsequently,
application
at
Company
W,
combined
with
ARM
(Association
Rule
Mining)
method,
revealed
inconsistencies
parameters.
After
proposing
validating
solutions
W’s
workshop,
findings
demonstrated
14.0%
reduction
occurrences
3.87%
enhancement
efficiency.
This
demonstrates
optimization
processes
manufacturing.
International Journal of Automation Technology,
Год журнала:
2024,
Номер
18(2), С. 302 - 315
Опубликована: Март 4, 2024
The
extraction
of
valuable
compounds
from
moringa
plants
involves
complex
processes
that
are
highly
dependent
on
various
environmental
and
operational
factors.
Monitoring
these
using
Internet
Things
(IoT)-based
multivariate
time
series
data
presents
a
unique
opportunity
for
improving
efficiency
quality
control.
Multivariate
data,
characterized
by
multiple
variables
recorded
over
time,
provides
insights
into
the
behavior,
interactions,
dependencies
among
different
components
within
system.
However,
with
increasing
complexity
volume
IoT
generated
during
extraction,
anomaly
detection
becomes
challenging.
objective
this
study
is
to
develop
robust
efficient
system
capable
automatically
detecting
anomalous
patterns
in
real
providing
early
warning
signals
operators,
facilitating
timely
interventions.
This
paper
proposes
novel
hybrid
unsupervised
model
combining
density-based
spatial
clustering
applications
noise
k
-nearest
neighbors
IoT-based
data.
We
conducted
extensive
experiments
real-world
demonstrating
effectiveness
practicality
our
proposed
approach.
In
comparison
other
methods,
method
has
highest
precision
value
0.89,
recall
accuracy
0.87.
Future
research
will
measure
optimize
actuators
(relays
motors)
action.
It
can
also
be
used
forecasting
algorithms
detect
anomalies
coming
minutes.
Sensors,
Год журнала:
2024,
Номер
24(19), С. 6414 - 6414
Опубликована: Окт. 3, 2024
The
Internet's
default
inter-domain
routing
system,
the
Border
Gateway
Protocol
(BGP),
remains
insecure.
Detection
techniques
are
dominated
by
approaches
that
involve
large
numbers
of
features,
parameters,
domain-specific
tuning,
and
training,
often
contributing
to
an
unacceptable
computational
cost.
Efforts
detect
anomalous
activity
in
BGP
have
been
almost
exclusively
focused
on
single
observable
monitoring
points
Autonomous
Systems
(ASs).
attacks
can
exploit
evade
these
limitations.
In
this
paper,
we
review
evaluate
categories
based
their
complexity.
Previously
identified
next-generation
detection
remain
incapable
detecting
advanced
those
designed
public
monitor
infrastructures.
Advanced
attack
requires
lightweight,
rapid
capabilities
with
capacity
quantify
group-level
multi-viewpoint
interactions,
dynamics,
information.
We
term
approach
anomaly
detection.
This
survey
evaluates
178
identifies
which
candidates
for
Preliminary
findings
from
exploratory
investigation
also
reported.
