Database Reference
In-Depth Information
of the underlying data in order to make diagnostic characterizations in
real time. Effective event-detection algorithms are required in order to
perform this task effectively.
The stock market often creates large volumes of data streams which
need to be analyzed in real time in order to make quick decisions about
actions to be taken. An example of such an approach is the MobiMine
approach [56] which monitors the stock market with the use of a PDA.
Such methods can be used for a wide variety of applications such as
knowing human movement trends [24], social image search [77], animal
trends [83] grocery bargain hunting [38], or more general methods for
connecting with other entities in a given neighborhood [82].
4.4 Environmental and Weather Data
Many satellites and other scientific instruments collect environmental
data such as cloud cover, wind speeds, humidity data and ocean currents.
Such data can be used to make predictions about long- and short-term
weather and climate changes. Such data can be especially massive if the
number of parameters measured are very large. The challenge is to be
able to combine these parameters in order to make timely and accurate
predictions about weather driven events. This is another application of
event detection techniques from massive streams of sensor data.
In particular, such methods have found numerous applications in pre-
diction of long-term climate change [40, 58, 67]. For example, one can
use various environmental parameters collected by sensors to predict
changes in sea surface temperatures, indicators specific to global warm-
ing, or the onset of storms and hurricanes. A detailed discussion on the
application of such methods for climate and weather prediction may be
found in [40].
5. Conclusions and Research Directions
Data streams are a computational challenge to data mining problems
because of the additional algorithmic constraints created by the large
volume of data. In addition, the problem of temporal locality leads to a
number of unique mining challenges in the data stream case. This chap-
ter provides an overview to the generic issues in processing data streams,
and the specific issues which arise with different mining algorithms.
While considerable research has already been performed in the data
stream area, there are numerous research directions which remain to
be explored. Most research in the stream area is focussed on the one
pass constraint, and generally does not deal effectively with the issue
of temporal locality. In the stream case, temporal locality remains an
