A Framework for Data Warehousing and Mining in Sensor Stream Application Domains - Evolving Application Domains of Data Warehousing and Mining - page 127

Window Size (AWS) approach, the linear interpo-
lation approach, the linear trend approach, and the
CARM approach (Jiang, 2007). All these methods
are applied to our proposed framework to answer
the user's request for missing sensor air temperature
value. We compared the estimation accuracy, run-
ning time and memory space usage when applying
each method to our proposed framework.
Our performance study shows that the proposed
domain-driven framework can be applied to the
environmental monitoring sensor stream applica-
tion for different data warehousing and mining
tasks. For the missing sensor value application
component in the sensor network application, the
closed pattern based association mining approach
is an area worth to explore.

Domain-driven data mining is an open research
field, especially for spatial, temporal sensor
stream data. We believe that in the future new
data warehousing and mining algorithms that
consider data semantics and domain information
have to be developed in order to extract more
interesting patterns and associations in different
application domains.

referenceS

Austin, F. I. (2000). Austin Freeway ITS Data
Archive . Retrieved January, 2003 from http://
austindata.tamu.edu/default.asp.

Cai, Y. D., Pape, G., Han, J., Welge, M., & Auvil,
L. (2004). MAIDS: Mining alarming incidents
from data streams. International Conference on
Management of Data .

concluSIon And
future trendS

Chang, J. H., Lee, & W. S. (2004). A sliding win-
dow method for finding recently frequent itemsets
over online data streams. Journal of Information
Science and Engineering .

Sensor stream applications are becoming very
common with the advances in technologies for
sensor devices. There data are normally available
as stream data with very little or no semantics. This
makes their analysis and knowledge extraction
very complex from an application point of view.
Most sensor stream mining works have focused
on the streaming properties without considering
the background of sensor domain information.
In this chapter we have addressed the problem
of mining sensor stream system from an applica-
tion point of view. We presented a framework
to preprocess sensor streams for domain-driven
data mining. The objective is to integrate domain
information that is relevant for stream data ware-
housing and data mining.
We have evaluated the framework with real
data from two different stream data application
domains, which shows that the framework is
general enough to be used in different application
scenarios. This is possible because the user can
choose the domain information that is important
for data warehousing and mining tasks.

Chang, J. H., Lee, W. S., & Zhou, A. (2003).
Finding recent frequent itemsets adaptively over
online data streams. ACM SIGKDD International
Conference on Knowledge Discovery and Data
Mining .

Chi, Y., Wang, H. X., Yu, P. S., & Muntz, R. R.
(2004). Moment: Maintaining closed frequent
itemsets over a stream sliding window. IEEE
International Conference on Data Mining .

Dang, X. H., Ng, W. K., & Ong, K. L. (2007).
Online mining of frequent sets in data streams
with error guarantee. Knowledge and Informa-
tion Systems .

Demaine, E. D., Ortiz, A. L., & Munro, J. I.
(2002). Frequency estimation of internet packet
streams with limited space. European Symposium
on Algorithms .