Today,
a
large
number
of
people
dabble
in
the
realm
social
media.
Due
to
pandemic
situation,
are
even
more
engaged
since
they
frequently
use
media
vent
their
emotions.
One
many
detrimental
effects
this
pervasive
usage
is
cyberbullying,
which
troubling
form
online
harassment.
Though
it
can
take
several
forms,
most
common
one
text.
Cyberbullying
on
media,
and
instead
confronting
perpetrator,
victims
often
have
mental
breakdowns
as
result
it.
This
study's
computerized
cyberbullying
detection
method
accesses
Twitter
users'
psychological
traits,
including
personalities,
moods,
Our
study
provides
an
innovative
solution
for
detecting
tweets
by
attention-based
transformer
algorithm
combined
with
embeddings.
model
acts
detector
classifying
that
related
cyberbullied
actions.
These
converted
into
numerical
vectors
Embeddings
divided
fixed
segments
through
padding
technique.
The
learns
from
encoder
part
comprising
self-attention
feed-forward
neural
network
normalization
tweet's
dataset.
Incredibly
accurate
made
possible
integrated
technology.
approach
promises
identify
quickly
precisely
give
control
women
over
situation.
International Journal of Computational and Experimental Science and Engineering,
Journal Year:
2025,
Volume and Issue:
11(1)
Published: Feb. 3, 2025
Depletion
of
dissolved
oxygen
in
the
water
is
a
serious
threat
to
fish
and
other
aquatic
organisms,
it
causes
aerobic
stress
disease
fish.
Detection
crucial
maintain
better
growth
spawning
fishes.
Recently
many
studies
proposed
deep
learning-based
quality
analysis
techniques,
but
these
techniques
inadequate
handling
complex
data.
Because
has
both
spatial
temporal
characteristics,
this
makes
most
learning
models
inadequate.
To
handle
such
multifaceted
data
we
ConvRec,
architecture
that
incorporates
CNN
(Convolution
neural
network)
LSTM
(Long-short
term
structures.
component
extracts
feature
domain
from
different
locations
while
captures
features
hence
model
can
learn
correlations
between
movement
parameters
classify
aqua
ponds.
In
work
use
two
dataset
are
unlabelled
collected
using
IoT
(Internet
things)
devices.
ConvRec
model,
usus
fine-grained
annotation
points
have
effect
empowering
detect
relevant
traits
associated
with
It
be
therefore
ascertained
yields
high
degrees
accuracy
99.2%
99.65%,
on
“ponds”
“waterx”
datasets
respectively
past
only
98.2%
98.1%
same
datasets.
These
results
demonstrate
not
promising
for
estimating
health
during
deficiency
also
take
part
reducing
negative
impact
low
levels
Electronics,
Journal Year:
2025,
Volume and Issue:
14(2), P. 331 - 331
Published: Jan. 15, 2025
The
aquatic
environment
in
aquaculture
serves
as
the
foundation
for
survival
and
growth
of
animals,
while
a
high-quality
water
is
necessary
condition
promoting
efficient
healthy
development.
To
effectively
guide
early
warnings
regulation
quality
aquaculture,
this
study
proposes
predictive
model
based
on
dual-channel
dual-attention
mechanism,
namely,
DAM-ResNet-LSTM
model.
This
encompasses
two
parallel
feature
extraction
channels:
residual
network
(ResNet)
long
short-term
memory
(LSTM),
with
mechanisms
integrated
into
each
channel
to
enhance
model’s
representation
capabilities.
Then,
proposed
trained,
validated,
tested
using
meteorological
parameter
data
collected
by
an
offshore
farm
environmental
monitoring
system.
results
demonstrate
that
structure
mechanism
can
significantly
improve
performance
prediction
accuracy
pH,
dissolved
oxygen
(DO),
salinity
(SAL)
(with
Nash
coefficients
0.9361,
0.9396,
0.9342,
respectively)
higher
than
chemical
demand
(COD),
ammonia
nitrogen
(NH3-N),
nitrite
(NO2−),
active
phosphate
(AP)
0.8578,
0.8542,
0.8372,
0.8294,
respectively).
Compared
single-channel
DA-ResNet
(ResNet
mechanism),
predicting
DO,
SAL,
COD,
NH3-N,
NO2−,
AP
increase
12.76%,
12.58%,
11.68%,
18.350%,
19.32%,
16%,
14.99%,
respectively.
DA-LSTM
(LSTM
corresponding
increases
are
9.15%,
9.93%,
9.11%,
10.91%,
10.11%,
10.39%,
10.2%,
ResNet-LSTM
LSTM
parallel)
without
attention
improvements
1.91%,
2.4%,
0.74%,
3.41%,
2.71%,
3.55%,
4.13%,
fulfills
practical
requirements
accurate
forecasting
nearshore
aquaculture.
Ecological Informatics,
Journal Year:
2023,
Volume and Issue:
79, P. 102405 - 102405
Published: Dec. 12, 2023
Water
contamination
presents
a
significant
challenge
in
aquaculture,
impacting
the
sustainability
of
ecosystems
and
health
aquatic
organisms.
Precisely
assessing
water
levels
is
crucial
for
effective
monitoring
safeguarding
life
within
aquaculture
industry.
Traditional
methods
evaluating
are
characterized
by
their
costliness,
time-consuming
nature,
susceptibility
to
errors.
Integrating
computer
technologies
such
as
Artificial
Intelligence
(AI),
Internet
Things
(IoT),
Data
Analytics
offers
promising
potential
addressing
this
issue.
Nevertheless,
current
deep
learning
solutions
have
limitations
related
data
variability,
interpretability,
performance.
To
address
these
limitations,
study
proposes
comprehensive
framework
that
incorporates
IoT-based
collection
segregation
techniques
enhance
accuracy
classification
aquaculture.
Real-time
collected
through
IoT
devices,
encompassing
parameters
like
temperature,
pH
levels,
dissolved
oxygen,
nitrate
concentration,
other
quality
indicators,
enables
holistic
evaluation
quality.
By
considering
predefined
acceptable
ranges
life,
calculates
index,
facilitating
into
categories
contaminated
non-contaminated.
ensure
robust
classification,
introduces
an
innovative
attention-based
model
known
Ordinary
Differential
Equation
Gated
Recurrent
Unit
(AODEGRU).
This
attention
mechanism
directs
model's
focus
towards
salient
features
associated
with
contamination,
while
AODEGRU
architecture
captures
temporal
patterns
data.
Experimental
results
underscore
effectiveness
proposed
model.
It
demonstrates
its
superiority
high
performance,
achieving
rate
approximately
98.69%
on
publicly
available
dataset
impressive
99.89%
real-time
dataset,
clearly
outperforming
existing
methodologies.
