Database Reference
In-Depth Information
fact, this step was not in the original process proposed by Fayyad, but was
introduced in the Athena system of
Baglioni et al.
(
2012
) where experiments on
two trajectories data sets representing cars and pedestrians were reported and
discussed.
The M-Atlas system has been introduced here as an example of a system
supporting the mobility knowledge discovery process with illustrating examples
along the chapter. This tool takes inspiration from the inductive database vision
by H.
Mannila
(
1997
) and it was originally introduced in
Giannotti et al.
(
2011
).
There, the experiments were run on two GPS data sets collecting car trajectories
from two Italian cities. Parts of these experiments are illustrated in Chapter
10
.
The implementation of the system is based on PostGIS spatial database system,
from which many of its spatial operators have been inherited.
The techniques presented in the preprocessing step derive from literature
works. For example, a survey on data set sampling techniques is presented in
the topic by
Scheaffer et al.
(
2005
) while progressive clustering on trajectory
data is introduced in the paper by
Rinzivillo et al.
(
2008
).
An approach for mobility understanding that has not been presented here
is data mining on semantic trajectories, represented as sequences of stops and
moves. In this case, standard data mining techniques such as frequent and
sequential patterns can be used. For example, in the work by
Alvares et al.
(
2007
), trajectories are first preprocessed to transform them into stop and moves,
which is essentially a relation representation as stated in
Spaccapietra et al.
(
2008
). Then, standard data mining techniques are applied. This simple but
clever technique allows the user to extract trajectory patterns that are purely
semantic and that cannot be found with classical spatio-temporal data mining,
based on the geometry of the trajectories. This has been the first approach to
facing the problem of mining semantic trajectories.
