Databases Reference
In-Depth Information
5 Discussion and Conclusion
The main reason for data mining to be developed more as and art than as a
science can be found in the lack of a methodology for data mining project de-
velopment. In this chapter, we have presented a first approach to such method-
ology and we have focused on the first phase, project conception, as the basis
for the methodology. A first step towards the systematization of data mining
project development is the definition of certain abstraction mechanisms to
capture the project goals as well as to define how to reach them. The objec-
tive of this abstraction is to provide the manager with a method to describe
goals in terms of data mining. This will help the later planning, managing
and developing processes involved in every project. Deeply analyzing the tar-
get domains of data mining tasks (structure, data generation processes and
data themselves) will help to identify shared concepts to every data mining
project no matter what the nature of the domain could be. This way it will
be possible to set the basis for defining elements to represent business goals
and therefore fully find answer, to questions such as: what is intended to do?,
which are the target elements to be analyzed?, what are the techniques to be
applied? or which requirements must fulfil the data?
In this chapter we have also presented a first approach to a global method-
ology for managing data mining projects based on RUP. Steps of the method-
ology have been established as well as deliverables to be obtained along the
process. Technical tasks of the methodology have been taken from CRISP-
DM. Nevertheless, to clearly define the methodology, deliverables have to be
precisely defined and an abstraction mechanism has to be found to express
deliverables in a standard way. On the other hand, activities related to the
management tasks have also to be defined.
Acknowledgments
The research has been partially supported by Ministerio de Educacion y Cien-
cia (project TIN2004-05873).
References
1. A. Chidanand and S. Weiss. Data mining with decision trees and decision rules.
Future Generation Computer Systems
, 13(2-3):197-210, 1997
2. R. Agrawal, H. Mannila, R. Srikant, H. Toivonen, and A. Verkamo. Fast discov-
ery of association rules.
In U. Fayyad et al. Advances in Knowledge Discovery
and Data Mining
. MIT, Cambridge, MA, 1996
3. American Society for Quality. Six Sigma Forum.
http://www.asq.org/info/
glossary/p.html
, last accessed 2005
4. M. Berry and G. Linoff.
Data Mining Techniques for Marketing, Sales and
Customer Support
. Wiley, New York, 1998
