Database Reference
In-Depth Information
challenges from the perspective of mining and analysis. These challenges
are as follows:
The volume of data associated with RFID can be extremely large,
because of the large number of tags which may be tracked by a
single reader. Furthermore, in some applications, the number of
readers may also be quite large, which leads to a
high fan-in and hi-
erarchical
organization of the underlying sensor network [23]. Such
a system poses numerous challenges, because successive hierarchi-
cal aggregation of streams from different nodes of the network can
lead to a overwhelming amount of data at the higher level nodes.
It has been argued in [23] that a uniform stream-oriented query
processing approach at all levels of the hierarchy works best in
such systems.
Thedatacanbevery
noisy
and
redundant
, with many tags being
completely dropped, and others being read by multiple readers
at multiple instants, resulting in tremendous redundancy of the
representation. Furthermore, the large volume of the data makes
the process of cleaning much more challenging. Therefore, effective
methods need to be designed to compress and clean such data.
The cleaning process can be rather expensive, and challenging,
especially when near real-time responses to location queries are
required.
Many applications such as high level semantic event detection can
be extremely challenging because of the high volume of the stream,
and the real time nature of such applications. The noise and errors
in the underlying data can lead to additional ambiguities during
the event detection process.
RFID deployments lead to a number of privacy concerns, because
tags are uniquely identifiable by readers. Therefore, by carrying
a tag attached to clothing, it may be possible to covertly track
people without their knowledge. A variety of methods need to be
designed in order to increase the privacy and security aspects of
RFID technology.
As with any sensor infrastructure, RFID technology and readers
vulnerable to partial or complete system failures. Such failures
can also lead to challenges in data processing, because data which
is not collected will always be missing from the database. If the
missing data is not explicitly accounted for by the underlying data
analytics, it may lead to inaccurate inferences, because the missing
