Database Reference
In-Depth Information
interesting locations, trajectories, or even the transportation modes of
the different users [180, 181].
While social sensing applications are generally defined for the case of
people, a similar analysis can be applied to the case of online tracking
of animals. For example, animals which are drawn from the same com-
munity or family may be considered to have implicit links among them.
Such links can be utilized for the perspective of detailed understanding
of how community and family membership affects geographical patterns.
Such information can be very useful for a variety of applications, such
as building disease propagation models among animals.
7.1 Integrating Sensor Data with Heterogeneous
Media for Enhanced Mining and Inference
Many of the devices (such as mobile phones), which enable social
sensing applications are
convergent devices
, which provide multiple func-
tionalities in recording different kinds of media data. For example, most
mobile phones today provide the capability to record photos, videos, text
blogging and tweets, and upload them directly in real time. Thus, such
media data automatically becomes
geo-tagged
, and this additional infor-
mation provides a rich source of information for improving the mining
process.
For example, the problem of providing location and activity recom-
mendations on the basis of user contributed comments and their GPS
trajectories has been studied in [179]. The user comments provide deeper
insights into their activity histories, which can be leveraged for a better
mining process. The collective wisdom of the trajectories and comments
of different users can be leveraged in order to provide answers to ques-
tions such as the following:
For a particular activity, what are the most appropriate places to
visit?
For a particular location, which a user has already visited, what
are the other activities that can be performed at that location?
In order to achieve this goal, the user location and activity histories
are used as the input. We note that the activity histories can only be
indirectly derived from user comments, by mining the relevant words
in the comments, which are related to specific activities. The location
features and activity-activity correlations are mined in order to obtain
additional knowledge. A collective matrix factorization methods was
applied in [179] in order to mine interesting locations and activities and
recommend them to users. Location information is also useful for rec-
