Database Reference
In-Depth Information
Figure 7.3 The flow of operations needed to extract the StuckInTrafficJam behavior.
of trajectory reconstruction cuts the movement of the users into trips describing
real activities and thus avoids long stops, for example, during the night or during
the working time.
CREATE DATA Trajectories AS MOVING_POINTS
FROM (SELECT t.userID, t.lon, t.lat, t.timestamp
FROM Observations)
SET MOVING_POINTS.MAX_TIME_GAP = 3 hours AND
MOVING_POINTS.MAX_SPACE_GAP = 50 meters
The result is a table called
Trajectories
with the three columns
userID
,
trajID
,and
trajectory
. We can notice that the trajectory becomes a data
type in the system. For space reasons we skip other possible preprocessing
steps and we proceed with the
data mining step
using the T-flock algorithm -
described in Chapter
6
- to obtain the groups of at least 10 cars with a maximal
distance of 20 meters between them for a time period of at least 5 minutes. These
parameters have been chosen as a reasonable approximation of what a candidate
traffic jam represents. However, this strongly depends on the application and the
characteristics of the analyzed urban area. Typically, bigger cities require larger
parameters for the flock to be identified as a traffic jam.
CREATE MODEL Flocks USING T-FLOCK
FROM (SELECT trajID, trajectory FROM Trajectories)
SET T-FLOCK.MIN_SUPPORT = 10 AND
T-FLOCK.MAX_SPACE_GAP = 20 meters AND
T-FLOCK.MIN_DURATION = 5 minutes
Once again, the result is stored in a table called
Flocks
. At this point we
have computed a spatio-temporal behavior, though we need to go a step further
toward a semantic behavior, like the StuckInTrafficJam. There are still a few
tasks to be done to identify traffic jams as the flocks with a low speed when the
