Next: Proof of Bayes Rule
Up: Conclusion and Further Work
Previous: Conclusion
There are several possible ways to continue the work on integration between
Case-Based Reasoning and Data Mining. It is possible to test other methods
of DM and CBR for integration, and to integrate in different ways:
- Test other Data Mining methods, like Inductive Logic Programming , and Rough Sets to produce rules that can be used as background knowledge.
One could also use informative methods like clustering with SPAD or
AutoClass to treat cases that belong to different classes differently.
- Compare results with other Case-Based Reasoning tools that are
knowledge rich like Creek [1] or that use other types of
similarity metrics than KATE.
- Investigate if it is possible to construct a Bayesian Network directly
from a clustering algorithm such as SPAD which will be used in NOEMIE to
produce the cases [26]. This might make the system more dynamic.
- Develop a tool-box for Data Mining/Case-Based Reasoning to make it easier to do experiments on a set of data with different similarity metrics and
Data Mining algorithms. Here the CRISP-DM initiative, mentioned in
section 4.2 could be used. Such a tool-box is available for
Data Mining [8], however without support for Bayesian
networks.
- Implement the CBRDM system further with routines for reuse ,
revise and retain .
- Research should also be done on the other methods for integration
which is outlined in chapter 4.
Next: Proof of Bayes Rule
Up: Conclusion and Further Work
Previous: Conclusion
Torgeir Dingsoyr
2/26/1998