Database Reference
In-Depth Information
approximately the same amount of time - because they simply share interests
and attitude. This leads to a new category of questions, which can be well
represented by the following:
Are there groups of objects that perform a sequence of movements, with similar
timings, though possibly during completely different moments?
Patterns such as flocks and moving clusters can provide some answers to the
question, but usually in small numbers, since the set of answers is limited to
movements that happen synchronously among all objects involved. The question
involves a much weaker constraint on the temporal dimension of the problem,
and therefore might allow many more answers. In the following we will present
one example of a pattern that goes in this direction and extracts spatio-temporal
behaviors that are followed by several objects, but allowing any arbitrary time
shift between them.
Trajectory patterns
(
T-Patterns
) are defined as sequences of spatial locations
with typical transition times, such as the following two:
Railway Station
15
min
−→
Museum
2
h
15
min
−→
Castle Square
Railway Station
10
min
−→
Middle Bridge
10
min
−→
Campus
For instance, the first pattern might represent the typical behavior of tourists
who rapidly reach a museum from the railway station and spend there about
two hours before getting to the adjacent square. The second pattern, instead,
might be related to students who reach the university campus from the station
by passing through the mandatory passage on the central bridge over the river.
A graphical example is also provided in Figure
6.3
a.
The two key points that characterize T-Patterns are the following: first, they
do not specify any particular route among two consecutive regions described:
instead, a typical travel time is specified, which approximates the (similar) travel
time of each individual trajectory represented by the pattern. In the gap between
two consecutive regions a trajectory might even have stopped in other regions
not described in the pattern. Second, the individual trajectories aggregated in
a pattern need not to be simultaneous, since the only requirement to join the
pattern is to visit the same sequence of places with similar transition times, even
if they start at different absolute times.
T-Patterns are parametric on three main parameters: the set of spatial regions
to be used to form patterns, that is, the spatial extension of “Railway Station”
and any other place considered relevant for the analysis
3
; the so-called
minimum
3
Actually, the algorithmic tool provided in literature to extract T-patterns also contains heuristics to
automatically define such regions, but in general the domain expert might want to do it manually in
order to better exploit knowledge or to better focus the analysis, or both.
