Database Reference
In-Depth Information
participants with the use of widely downloadable mobile applications, as
long as appropriate privacy-protection mechanisms are in place. A clas-
sic example of such a data set is the well known
GeoLife
data set [194].
Such data sets are not just collected for humans, but even from animals
for tracking purposes. An example of such an animal tracking database
is the
Movebank
database [186], which contains detailed data about an-
imal trajectories in the data. Finally, many recent document and image
creation hardware such as GPS-enabled cameras and cellphones auto-
matically stamp the content with GPS locations. This creates a very
rich data set containing
both
content and (implicit) trajectories. The
availability of such data makes it important to design more effective and
ecient methods for trajectory mining.
Trajectory data is particularly useful from the perspective of mining
aggregate community movement patterns. A variety of interesting pat-
terns can be mined in such trajectory data sets, which provide insights
into the aggregate movements. The aggregate movements are best rep-
resented by clusters, which are variously referred to as
flocks
,
convoys
,
or
swarms
[22, 77, 91, 102, 104, 113], depending upon the model which
is used to characterize these clusters. Typically, the goal is to either
determine objects with trajectories of similar shape, or objects which
move together in clusters. The major difference between these different
kinds of moving clusters are as follows:
Flocks:
These correspond to groups of objects which move within
a fixed disc of a particular size over consecutive time-stamps [77,
102]. As a result, the underlying trajectories will often have a
similar geometric shape.
Moving Cluster:
This refers to a group of objects which have con-
siderable overlap between successive time-stamps [93]. As in the
previous cases, the constraint on the objects moving together in
successive time stamps leads to trajectories of similar shape.
Convoys:
In this case, we again find groups of objects which move
together, except that the concept of density is used in order to
define the objects that move together. As before, the objects need
to move together over consecutive time stamps [90, 91]. In many
scenarios, the use of density provides a flexible way of modeling
the movement of significant masses of objects together.
Swarms:
In the case of swarms, the objects are required to move
together as before, except that we do not impose the requirement
that the objects should be together over consecutive time stamps
