Geography Reference
In-Depth Information
introduced. Explanations of the field study and data collection will follow, and the
subsequent section reports the application of the algorithm and explains as well as
discusses the results of trip purpose detection. The last section concludes the paper.
13.2
Methods for Detecting Trip Purposes from GPS Data
In order to derive activity types and trip purposes from GPS data, it is generally
agreed that some additional information such as land use and point of interests
(POIs, e.g., public facilities, shopping centers, and educational establishments) are
needed (Bohte and Maat
2008
;Wolf
2000
). In addition, data like respondents'
home address, work place and places that they frequently visit may also be required
(Stopher et al.
2008
;Wolfetal.
2004
).
Wo l f (
2000
) demonstrates that trip purposes might be detected by matching the
land use types of trip ends with the possible trip purposes. A table of standardized
land use types and corresponding trip purposes is developed. A primary trip purpose
is identified for each land use description and two additional possible trip purposes
are added to accommodate the cases that different types of trip purposes may be
associated with the same type of land use. In a follow-up study, Wolf et al. (
2004
)
propose to collect data on respondents' socio-demographics and the destinations of
their habitual activities to supplement GPS and GIS data for detecting trip purposes.
Some rules are developed. For example, trip ends within a radius of 200 m from
home are assigned with the purpose of 'Home'. Other trip purposes were derived
by matching the GPS data with available POI information and land use polygons
in GIS. Stopher et al. (
2008
) suggest that information on workplace, home address
and the location of the often visited grocery stores can help ascertain about 75 % of
trip ends and over 60 % of trip purposes. As for the remaining trip purposes, some
heuristic rules may be applied to deduce with the help of land use data and data
on occupation. Stopher and Collins (
2005
) advocate using internet-based prompted
recall survey by showing respondents the derived results and asking for additional
information to improve trip identification and trip purpose detection. Similarly,
Bohte and Maat (
2008
) make use of the Internet to help validate trips and trip
purposes identified from GPS data. A user friendly web-based interface is developed
to visualize the trajectories of movement and present trip information derived from
GPS data to respondents. Respondents can adjust the information about the trips
derived and add any missing trips or other information. Except for some heuristic
rules, most of the studies reviewed so far did not employ any model or algorithm for
processing the data to derive trip purposes.
Griffin and Huang (
2005
) propose to use a decision tree classification model to
derive trip purpose. Trip characteristics such as spatial location, time of day, length
of stay are used as attributes to build the decision tree, which leads to the classifica-
tion of different trip purposes. Similarly, Deng and Ji (
2010
) suggest establishing an
initial decision tree for purpose classification based on attributes like land-use and
temporal information, duration of stay and socioeconomic characteristics. Instead
Search WWH ::
Custom Search