Geography Reference
In-Depth Information
adopted an interactive validation approach by providing a user friendly interface
consisting of a map and a table that depict all trips that a respondent visited during a
day and present to the respondent for verification. The interactive approach requires
substantial involvement of respondents and thus may not be able to overcome the
low response rate problem of traditional data collection methods. On the other
hand, as the name implies, passive GPS does not require the active involvement of
respondents. Respondents are passively involved in the survey by simply carrying
the device along with them and turn it on/off in time without the need to input any
travel-activity behavior data. As a result, burdens on respondents can be greatly
reduced. However, passive GPS data includes information only on time, position
as well as heading and speed of movement etc. at certain time intervals. Most of
the activity-travel information, such as what (types of activities and trip purposes),
where (place where activities are conducted as well as trip origins and destinations),
when (starting time, duration as well as trip arrival and departure time) and what
transportation modes are used, have to be detected. Among the most difficult travel
behavior information to be derived from GPS data is that on activity types and trip
purposes.
In the past years, many attempts have been made to derive activity-travel infor-
mation from GPS data. For examples, Wolf et al. ( 2004 ) proposed and developed
methods to derive from GPS data trip purpose information with the help of GIS data
on public facilities, shopping centers, residential places, etc. A number of similar
attempts in combining GPS tracking data with GIS land use data and analysis
to detect transport mode and trip purposes have been reported in the literature
(Chung and Shalaby 2005 ; Bohte and Maat 2008 ;Chenetal. 2010 ). While the
GIS functionalities such as buffer and overlay and simple rules of thumb are used
in these studies to derive from GPS data the needed information, in most cases no
explicit methodologies are applied. The only exception is Chen et al. ( 2010 ), who
employed a probability model to determine the probability that a trip had one of
the predefined trip purposes. Recently, methods with self-learning capabilities have
been proposed to mine GPS tracking data and detect information such as transport
modes and trip purposes. Moiseeva et al. ( 2010 ) suggested combining Bayesian
belief network with self-learning to detect trip mode and purpose. Activity-travel
patterns that are confirmed or corrected by respondents are used to improve the
conditional probabilities for classification. Deng and Ji ( 2010 ) proposed using a
decision tree algorithm to derive trip purposes.
This study suggests applying a genetic algorithm, an artificial intelligent method,
to identify trip purpose through mining GPS tracking, land use and socio-economic
data. Specifically, this method establishes a set of classification models through self-
learning to extract the information of trip purpose based on information concerning
land use type of trip ends, trip duration and timing, socioeconomic characteristics
of respondents, and other relevant information. The study seeks to contribute to
recent attempts to pave the way towards the development of an entirely passive
and automatic way of collecting data from individuals for activity-travel behavior
studies. The next section will review the existing methods for trip purpose detection.
Then the genetic algorithm and its application for trip purpose detection will be
Search WWH ::




Custom Search