Database Reference
In-Depth Information
rumors or other information in the community. In a static network such
as
Facebook
, the problem of influence analysis is much more straight-
forward, because it depends upon the static connections between the
different communities [96]. In a dynamic network, the underlying net-
work structure may change rapidly over time, depending upon the in-
teractions between the underlying entities. Some recent work on dy-
namic influence analysis addresses this scenario of interactions between
dynamic and evolving entities [11]. This method can determine either
influential nodes or determine the most likely points of release, based
on a given influence pattern and also a given pattern of interactions. A
classic example of a dynamic network in the context of social sensing
is the
face-to-face interaction network
, in which it may be desirable to
determine the influence of such interactions on specific behaviors. For
example, the work in [122] used a mobile phone-based sensing platform to
examine the influence of face-to-face interactions in the life-style choices
of participants such as obesity, eating and exercise habits. It was shown
that the use of sensing platforms can be very effective at modeling the
influence effects of such interactions (which turned out to be significant
for this scenario).
As discussed earlier, the determination of dynamic interactions can
sometimes require the real-time modeling of
implied interactions
(such
as face-to-face interactions), which are hard to infer from sensor data
can also sometimes be sensitive information. This also leads to nu-
merous privacy challenges, especially since the interactions between the
participants may be considered personal information. As mentioned ear-
lier, privacy continues to be an important issue for such social sensing
applications. A number of privacy-sensitive approaches for face-to-face
activity modeling and conversation segmentation have been discussed in
[164-167].
The dynamic modeling of social sensing applications, naturally lead
to a lot of trajectory data in real applications. Therefore, significant
amount of research has been devoted to determining spatio-temporal
patterns from such trajectories. Such patterns may be derived with or
without additional content information. A number of these methods will
be discussed in the next section.
7. Trajectory Mining for Social Sensing
Social sensing applications have naturally lead to the collection of tra-
jectory database from the rich GPS data, which is collected in a wide
variety of applications. The increasing popularity and availability of
mobile phones also enables the collection of trajectory data from willing
