Electronic
medical
record
mining
based
on
relationship
classification
has
become
a
hot
topic
in
the
field
of
healthcare.
However,
existing
models
classification,
most
them
use
single-layer
attention,
it
results
relatively
simple
feature
representation
and
is
easy
to
lose
information
during
training.
Therefore,
this
paper
proposes
method
dual
channel
attention.
Firstly,
1
combines
BERT(Bidirectional
Encoder
Representation
from
Transformers),
GRU(Gate
Recurrent
Unit)
Global
Attention,
while
2
Subject_object_mask_generation
So
Attention.
Specifically,
we
module
specify
corresponding
positions
subject
object
within
text.
And
Attention
used
focus
attention
between
object.
Secondly,
outputs
two
channels
are
concatenated.
Finally,
perform
concatenated
results.
We
evaluated
public
dataset
CMeIE(Chinese
Medical
Information
Extraction),
experimental
showed
that
improved
model's
accuracy,
recall
F1
values
increased
by
2.2%,
0.03%
1.3%
respectively,
compared
baseline.
It
indicates
our
certain
advantages
other
methods.
Energy and AI,
Journal Year:
2024,
Volume and Issue:
16, P. 100371 - 100371
Published: April 17, 2024
This
paper
proposes
an
integration
of
recent
metaheuristic
algorithm
namely
Evolutionary
Mating
Algorithm
(EMA)
in
optimizing
the
weights
and
biases
deep
neural
networks
(DNN)
for
forecasting
solar
power
generation.
The
study
employs
a
Feed
Forward
Neural
Network
(FFNN)
to
forecast
AC
output
using
real
plant
measurements
spanning
34-day
period,
recorded
at
15-minute
intervals.
intricate
nonlinear
relationship
between
irradiation,
ambient
temperature,
module
temperature
is
captured
accurate
prediction.
Additionally,
conducts
comprehensive
comparison
with
established
algorithms,
including
Differential
Evolution
(DE-DNN),
Barnacles
Optimizer
(BMO-DNN),
Particle
Swarm
Optimization
(PSO-DNN),
Harmony
Search
(HSA-DNN),
DNN
Adaptive
Moment
Estimation
optimizer
(ADAM)
Nonlinear
AutoRegressive
eXogenous
inputs
(NARX).
experimental
results
distinctly
highlight
exceptional
performance
EMA-DNN
by
attaining
lowest
Root
Mean
Squared
Error
(RMSE)
during
testing.
contribution
not
only
advances
methodologies
but
also
underscores
potential
merging
algorithms
contemporary
improved
accuracy
reliability.
Journal of Reliable Intelligent Environments,
Journal Year:
2024,
Volume and Issue:
10(3), P. 299 - 318
Published: July 26, 2024
Abstract
Amid
the
rise
of
mobile
technologies
and
Location-Based
Social
Networks
(LBSNs),
there’s
an
escalating
demand
for
personalized
Point-of-Interest
(POI)
recommendations.
Especially
pivotal
in
smart
cities,
these
systems
aim
to
enhance
user
experiences
by
offering
location
recommendations
tailored
past
check-ins
visited
POIs.
Distinguishing
itself
from
traditional
POI
recommendations,
next
approach
emphasizes
predicting
immediate
subsequent
location,
factoring
both
geographical
attributes
temporal
patterns.
This
approach,
while
promising,
faces
with
challenges
like
capturing
evolving
preferences
navigating
data
biases.
The
introduction
Graph
Neural
(GNNs)
brings
forth
a
transformative
solution,
particularly
their
ability
capture
high-order
dependencies
between
POIs,
understanding
deeper
relationships
patterns
beyond
connections.
survey
presents
comprehensive
exploration
GNN-based
recommendation
approaches,
delving
into
unique
characteristics,
inherent
challenges,
potential
avenues
future
research.
BIO Web of Conferences,
Journal Year:
2025,
Volume and Issue:
163, P. 01003 - 01003
Published: Jan. 1, 2025
In
the
rapidly
advancing
field
of
cancer
genomics,
identifying
new
genes
and
understanding
their
molecular
mechanisms
are
essential
for
targeted
therapies
improving
patient
outcomes.
This
study
explores
capability
Graph
Convolutional
Networks
(GCNs)
integrating
complex
multiomics
data
to
uncover
intricate
biological
relationships.
However,
inherent
complexity
GCNs
often
limits
interpretability,
posing
challenges
practical
applications
in
clinical
settings.
To
enhance
explainability,
we
systematically
compare
two
state-of-the-art
interpretability
methods:
Integrated
Gradients
(IG)
SHapley
Additive
exPlanations
(SHAP).
We
quantify
model
performance
through
various
metrics,
achieving
an
accuracy
76%
Area
Under
ROC
curve
is
0.78,
indicating
model’s
effective
identification
both
overall
predictions
positive
instances.
analyze
explanations
provided
by
IG
SHAP
gain
more
knowledge
decision-making
processes
GCNs.
Our
framework
interpret
contributions
omics
features
GCN
models,
with
highest
score
observed
feature
MF:UCEC
KIF11.
approach
identifies
novel
clarifies
mechanisms,
enhancing
interpretability.
The
improves
accessibility
personalized
medicine
contributes
biology.
