Database Reference
In-Depth Information
the mapping file, or can belong to the class “HomeToWork” as inferred by the
deductive step by the ontology engine. Therefore, the query:
SELECT t.id, t.trajectory
FROM trajectories t
WHERE 'HomeToWork behavior' in SEMANTIC(t.trajectory)
returns all the trajectories classified as HomeToWork by the deductive step
based on the axioms definition and the ontology inference engine. Now we can
combine the results with the concepts of
inside area
and
outside area
of the city
contained in a table called
Areas
to extract the CommuterMovement. This is
depicted by the following query:
CREATE TABLE CommuterMovement AS
SELECT t.trajId, t.trajectory
FROM trajectories t, areas a, areas a2
WHERE 'HomeToWork behavior' in SEMANTIC(t.trajectory)
AND 'Systematic movement' in SEMANTIC(t.trajectory)
AND ST_contains(ST_PointN(trajectory,first), a.area)
AND ST_contains(ST_PointN(trajectory,last), a2.area)
AND a.label = 'outside' and a2.label = 'inside'
where
ST_contains
3
is a spatial predicate which checks if a point is contained
in a specific area. In Figure
7.6
we show an example of the resulting trajectories.
Similarly to the TrafficJam case presented above, we can combine the results of
these two analyses, obtaining the commuters who are stuck in a traffic jam.
SELECT sj.userID, sj.flock
FROM CommuterMovement cm, StuckInTrafficJam sj
WHERE cm.trajID = sj.trajID
7.4 Conclusions
This chapter introduced a step-by-step KDD process for mobility understanding
by using examples from the use of the M-Atlas system. We have shown that
the understanding of mobility data is a complex process that involves many
different steps, all of which are necessary for the proper understanding of
the mobility phenomena. These steps are presented here explaining the tech-
niques that have to be applied to find meaningful behaviors. During this process
we observe an increasing involvement of semantic and contextual information
embedded progressively into the process. We have defined, as the final result
3
This function and the other used in the query derive from PostGIS and they can be used directly in
the M-Atlas system.
