Journal of Green Building,
Год журнала:
2025,
Номер
20(2), С. 55 - 76
Опубликована: Май 1, 2025
ABSTRACT
This
research
investigated
the
key
factors
that
influenced
patients’
individual
thermal
sensations
in
a
rehabilitation
ward.
Maintaining
comfort
is
important
for
occupant's
well-being
healthcare
facilities.
The
commonly
used
Predicted
Mean
Vote
(PMV)
model
has
limitations
on
considering
an
individual's
needs,
especially
if
impaired
health.
There
was
lack
of
sensation
studies
medical
settings.
study
conducted
ten-week
fieldwork
real
environment
order
to
develop
analysis
could
help
understand
patient's
needs.
Traditional
statistical
models
and
artificial
neural
network
(ANN)-based
models,
using
real-world
data
including
spatial
healthcare-related
parameters,
were
established
comparative
study.
results
unveiled
substantial
influence
parameters
inpatients’
indoor
sensations.
Furthermore,
ANN-based
demonstrated
better
performance
aligning
with
conditions
providing
more
accurate
prediction
outcomes
compared
traditional
model.
These
findings
can
be
by
hospital
designers
engineers
optimize
overall
quality
within
environment.
Smart and Sustainable Built Environment,
Год журнала:
2025,
Номер
unknown
Опубликована: Фев. 8, 2025
Purpose
By
harnessing
technology
developments
such
as
Internet-of-Things
(IoT)-enabled
intelligent
sensors
and
immersive
virtual
reality
(VR)
experiences,
facility
managers
can
access
real-time,
precise
information
on
thermal
comfort-related
indicators
through
models.
While
prior
research
studies
have
developed
key
technologies
for
improving
the
understanding
of
comfort
its
impact
occupants’
well-being
productivity,
there
remain
areas
yet
to
be
explored,
especially
in
relation
integrating
both
real-time
data
from
multimodal
IoT-enabled
smart
VR
technologies.
Hence,
this
study
demonstrates
potential
IoT
assessment
visualization
well
user
interaction
with
HVAC
systems
enhance
comfort.
Design/methodology/approach
To
develop
proposed
integrated
analytical
framework
paper,
various
steps
were
implemented.
First,
four
sensing
stations
created
installed
collect
(i.e.
temperature
relative
humidity).
Second,
a
environment
was
using
Unity
engine
offer
an
experience.
Third,
into
by
transmitting
it
cloud
via
MQTT
protocol
server,
programming
scripts
provide
multiple
functionalities
users,
including
visualizing
along
entire
indoor
space
interacting
controlling
cooling
heating
systems.
Fourth,
applicability
effectiveness
validated
evaluated
92
participants
survey
questionnaire.
Findings
The
obtained
results
importance
aspects
graphical
satisfaction,
spatial
presence,
involvement,
experienced
realism,
low-to-no
cybersickness
overall
application
among
others.
More
specifically,
findings
reflected
that
participants’
average
scores
sense
involvement
realism
4.69,
4.61,
4.71
4.53
out
5,
respectively.
showed
capabilities
serve
powerful
feature
enables
comprehensive
variations
across
room/office
space.
Also,
no
statistically
significant
differences
between
responses
experience
those
limited-to-no
experience,
thus
further
highlighting
usefulness
not
only
users
but
also
different
skills
background.
Originality/value
This
has
revolutionize
way
built
environments
are
managed
interacted
with,
where
monitor,
assess
visualize
interact
control
devices
real-world
distance
framework’s
ability
dynamic
continuously
updated
assessments
conditions
positions
valuable
tool
prompt
adjustments
optimize
levels.
Ultimately,
provides
intuitive
platform
manage
comfort,
promoting
healthier,
more
productive
eco-friendly
environments.
IEEE Access,
Год журнала:
2024,
Номер
12, С. 30039 - 30053
Опубликована: Янв. 1, 2024
Occupants'
personal
thermal
comfort
(PTC)
is
indispensable
for
their
well-being,
physical
and
mental
health,
work
efficiency.
The
heating
system
controlled
by
Artificial
intelligence
(AI)
can
calibrate
the
indoor
condition
automatically
analyzing
different
physiological
environmental
variables.
Predicting
occupants'
preferences
in
a
smart
home
be
prerequisite
to
adjusting
temperature
that
might
provide
comfortable
environment.
Modeling
preference
challenging
due
two
major
challenges
including
inadequancy
of
data
it's
high
dimensionality.
Adequate
amount
an
obvious
requirement
training
efficient
machine
learning
(ML)
models.
In
addition,
high-dimensional
tends
have
multiple
features
are
irrelevant,
noisy
hinder
ML
models'
performance.
To
this
end,
we
proposed
prediction
framework
employing
synthetically
generated
introducing
generative
adversarial
network
(CTGAN)
feature
selection
techniques.
We
first
address
inadequacy
challenge
applying
CTGAN
generate
synthetic
which
considers
associated
with
multimodal
distributions
categorical
features.
Then
techniques
incorporated
identify
best
possible
sets
from
sets.
wide
range
experimental
settings
on
standard
dataset
demonstrated
state-of-the-art
performance
predicting
preference.
results
clearly
indicate
models
trained
achieved
significantly
better
than
real
data.
turn,
our
methods
supervised
higher
terms
evaluation
metrics
accuracy,
Cohen's
kappa,
area
under
curve
(AUC)
outperforming
conventional
methods.
Additionally,
method
enhances
explainability
provides
avenue
experiment
designers
consciously
select
collected.
