Next: Further Work
Up: Conclusion and Further Work
Previous: Conclusion and Further Work
From the discussion in chapter 6, the results in in
chapter 5 and the theory in the previous chapter, we can draw
the following conclusions:
- The results from the CBRDM tool developed and KATE, differ greatly
in the features that are left blank in the ``new case'' under retrieval.
- The number of features and feature values must be reduced in order
to produce a Bayesian network with Bayesian Knowledge Discoverer in
reasonable time.
- The integration of Bayesian networks and Case-Based Reasoning can
lead to better results if the networks represents sound knowledge. This
improvement is better when a large number of features are given, and when
the given features are sub-nodes in the Bayesian network.
Torgeir Dingsoyr
2/26/1998