Transportation Research Record Journal of the Transportation Research Board,
Год журнала:
2024,
Номер
unknown
Опубликована: Март 30, 2024
While
existing
literature
has
extensively
explored
the
impact
of
online
shopping
on
travel
behavior,
few
studies
have
undertaken
segmentation
analysis
to
uncover
hidden
behavioral
heterogeneity.
This
study
fills
this
gap
by
addressing
heterogeneity
and
identifying
distinct
shopper
segments
based
behaviors,
with
a
focus
product
types.
Data
collected
in
November
December
2021
from
1,747
shoppers
Florida
were
analyzed
using
Latent
Class
Analysis
(LCA)
covariates.
Sociodemographic
residential
characteristics,
COVID-19
influences,
attitudes,
perceptions
channel-specific
factors
served
as
active
inactive
covariates
predict
class
membership.
Our
model
identified
six
classes
shoppers,
short-distance
dual-channel
representing
largest
(28.4%)
exclusive
smallest
(6.2%).
Dual-channel
shopaholics,
overrepresented
Gen
Zers,
Millennials,
Blacks,
workers,
exhibited
high
average
monthly
vehicle
miles
traveled
(VMT)
across
all
types
strong
potential
for
complementary
behavior.
Conversely,
members
silent
generation,
those
who
live
alone,
no
vehicle,
do
not
enjoy
shopping,
demonstrated
substitutive
In
general,
single-channel
showed
lower
VMT
than
their
counterparts
These
findings
contribute
deeper
understanding
offering
insights
more
accurate
quantification
net
traffic
environmental
impacts
e-commerce.
Additionally,
they
provide
valuable
considerations
designing
segment-specific
policies
aimed
at
minimizing
maximizing
shopping.
Traffic Injury Prevention,
Год журнала:
2025,
Номер
unknown, С. 1 - 9
Опубликована: Апрель 4, 2025
Crash
pattern
recognition
and
characterization
are
essential
for
reducing
the
damage
vulnerable
road
users
(VRUs)
suffer
in
motor
vehicle
crashes.
However,
traditional
methods
provide
an
incomprehensive
understanding
of
crash
causality
impacts
VRU-vehicle
interactions.
Therefore,
this
study
aims
to
a
reasonable
various
types
To
achieve
goal,
three-layer
causal
analysis
framework
was
developed.
The
layers
consist
physical
states
(mainly
environmental
human
factors),
interactions
(pre-crash
behaviors
drivers
VRUs),
First,
latent
class
cluster
sequence
were
used
identify
interactive
behavior
patterns
pairs,
respectively.
Besides,
oversampling
algorithm
proposed
assist
Granger
test
uncovering
relationships
between
pre-crash
patterns.
Finally,
Sankey
diagrams
utilized
compare
analyze
path.
results
show
that
single
consecutive
crashes
have
nine
eleven
typical
scenarios,
respectively,
excluding
considering
potential
chains.
These
chains
new
scenarios.
It
found
personal
subjective
factors
primarily
influence
drivers,
while
VRUs,
traffic
environment
plays
crucial
role.
Noteworthily,
highest
risk
only
associated
with
chain
where
vehicles
unable
brake
time.
Clarifying
interaction
is
essential,
which
can
help
finding
critical
causes
fatal
identified
VRU
violations
inability
time
as
determinants
severity
both
Accordingly,
targeted
safety
interventions
proposed,
including
enhancements
pedestrian
crossing
infrastructure
improvements
braking
systems
mitigate
risk.
Heliyon,
Год журнала:
2023,
Номер
9(10), С. e20468 - e20468
Опубликована: Сен. 28, 2023
Indicator
diagram
is
the
key
basis
for
fault
diagnosis
of
pumping
wells
in
oil
exploitation.
With
rapid
development
machine
learning,
indicator
based
on
deep
learning
has
garnered
increasing
attention.
This
kind
methods
train
neural
network
models
with
marked
samples,
and
then
inputs
images
into
trained
outputs
their
categories.
At
present,
preparation
sample
set
relies
experts'
analysis
one
by
one.
However,
it
involves
extensive
manual
work
marking
prone
to
errors,
so
samples
are
often
insufficient
quantity.
In
order
quickly
mark
a
large
number
well
data
was
plotted
standardized
diagram,
three
feature
extraction
diagrams
were
proposed:
original
vector,
three-dimensional
pixel
tensor,
convolutional
network.
These
convert
corresponding
vectors,
which
clustered
using
K-means
clustering
algorithm,
enabling
be
classified
different
categories
results.
Using
20,000
randomly
selected
pieces
from
100
wells,
this
study
clusters
proposed
methods.
The
results
indicated
that
time
consumption
0.2,
8.3,
0.7
h,
accuracy
rates
98%,
92%,
95%,
respectively.
For
diagrams,
method
vector
outstanding
performance
terms
efficiency
accuracy.
provides
an
automatic
tool
dataset,
its
can
increased
tens
times
compared
marking.
