Database Reference
In-Depth Information
7
Understanding Human Mobility
Using Mobility Data Mining
Chiara Renso and Roberto Trasarti
7.1 The Mobility Knowledge Discovery Process
We often say that “Knowledge is power” and this is particularly true in the
mobility field, because mobility knowledge gives great power in terms of appro-
priate decision making to actors in several application files, ranging from traffic
management to urban planning to ethology, just to give some examples. Mobility
knowledge can be rephrased as
how, when, where, and why objects move?
For
example, a traffic manager could improve traffic sustainability in a city when
he or she discovers why a specific traffic congestion happens, or an ethologist
could finally gain a deep understanding of why a given animal behaves in a given
way.
The mobility data mining research field has seen a growing interest in the
last few year - as already stated in Chapter
6
- providing several algorithms and
techniques tailored on trajectory data. However, a common problem of these
techniques is that the knowledge produced by the mining step is generally not
really applicable to the application domain as it is. The lack of semantics of the
extracted patterns makes the interpretation task far from obvious. This problem,
commonly recognized in the data mining literature, is particularly significant in
mobility data mining, where the complexity of the data themselves, together with
the extreme multifariousness of the mobility application requirements, makes
the knowledge discovery even more challenging. The interestingness evaluation
of the discovered movement patterns makes sometimes the KDD process a mere
academic application, useless in reality. A way to close the gap between “KDD
knowledge” and “application actionable knowledge” is to reinforce the KDD
process with tools capable of easily managing the steps integrated with new
elements focussed on the semantic aspect for an improved data and patterns
understanding. This results in a number of techniques, tailored to mobility data,
127
