Database Reference
In-Depth Information
(a) (b) (c)
Figure 7.4 A graphical representation of the process of extracting traffic jams from the
data. (a) Using the T-flock algorithm all the candidates are extracted. (b) The patterns are
colored based on ratio between their speed and the free speed in the same area (Blue
>
1,
Red
<
1). (c) The patterns with a speed lower than 1/4 of the free speed. (See color plate.)
Once we have the traffic jams, we can retrieve the trajectories of the users
who are stuck there using the a
data-model manipulation
realized through the
Entail
relation predicate:
CREATE RELATION StuckInTrafficJam USING ENTAIL
FROM (SELECT flockID, flock FROM TrafficJams),
(SELECT userID, trajID, trajectory)
The obtained table contains the set of trajectories of the users who are part
of a traffic jam (identified by
flockID
). In Figure
7.4
we visualize some of
the steps on the map. However, it is important to notice how this process does
not complete the understanding of mobility. In fact, the selected trajectories
can be further analyzed to determine, for example, the reasons of the traffic
jams. An example is to combine the StuckInTrafficJam with the Commuter-
Movement to discover a possible relation of a traffic jam with the commuting
behavior.
We have seen in the previous example how the semantic information is
embedded into the discovery process when passing from a spatio-temporal
behavior to a semantic behavior, for example, passing from the flocks to the
StuckInTraffic behavior. We have used domain information in the M-Atlas
queries to identify the semantic behavior from the extracted flocks. However,
the semantic enrichment step is not explicit in the process and it is somehow
embedded into the M-Atlas queries by the analyst. A further step in the direc-
tion of extrapolating and modularizing the semantic enrichment task from the
KDD process is to define the KDD process as a combination of induction (or
mining) and deduction (inference of a semantic behavior) reasoning tasks. The
framework Athena offers a solution: an extension of M-Atlas exploiting the
integration of ontologies in the mobility knowledge discovery process. Essen-
tially, this new process consists of a querying and mining process enhanced with
