A Framework for Data Warehousing and Mining in Sensor Stream Application Domains - Evolving Application Domains of Data Warehousing and Mining - page 120

Figure 4. Architecture of a domain-driven framework for sensor stream application

tion and constraints, such as data description, data
range, data type, etc.
The second level is the preprocessing tasks,
where the raw data are preprocessed and integrated
with the domain information, and the sensor iden-
tifier are connected with sensor values. For data
warehousing and mining, another fundamental
task is the transformation of the data in different
granularity levels. This is essentially important for
sensor stream data, where both space and time need
to be aggregated in order to allow the discovery
of patterns. The user may aggregate both space
and time in different granularities.
On the third level of the framework are the
data warehousing and data mining tasks. Once
the raw data are preprocessed and transformed,
different data mining tasks can be applied. For
example, data classification, data clustering,
pattern and association discovering, and online
data analytic processing, etc. These tasks can be
combined together to get different mining results
requested by the end users.

On the top of the framework is the query pro-
cessing component, where the end users select
query parameters and their queries can be answered
based on the mining results. Different end users'
requests can be answered at the same time accord-
ing to their specified query parameters.

cASe Study

In this section we describe two different appli-
cations and present some experiments to show
the usability of the framework for sensor stream
application domain. Our experiments were per-
formed with different data warehousing and data
mining methods.

the traffic Management
Sensor Stream Application

The simulation data of the traffic management
application was collected in year 2000 at various