If we place the CBRDM method in the framework of different integration approaches outlined in chapter 4, the best fitting category would be a method that uses Data Mining to obtain domain knowledge for use in a Case-Based Reasoning system. It is thus a master-slave system with CBR as the master. We use the dependency modeling technique Bayesian networks to find domain knowledge from a casebase. This knowledge is informative , where we can discover dependencies. This knowledge can be edited by an expert before using it in the CBR system, for example by using the Microsoft Belief Networks tool. The CBR system which is developed is extremely simple, and can be said to be knowledge poor . A knowledge rich method would be expected to have some deeper understanding of the data and not just use a network structure for inference.
When comparing the method to existing approaches, it has some similarity with the general DM-CBR systems developed at Rostock and NEC in that it uses Data Mining on the case data to infer domain knowledge. But it is perhaps more similar to the ``Aladdin'' system developed at Microsoft Research in that it uses a Bayesian network for inference. But there are two major differences between the new system and earlier approaches. CBRDM constructs the Bayesian network automatically from a casebase, while ``Aladdin'' requires an expert to design the network. Also, CBRDM makes use of the network in retrieval of cases, while ``Aladdin'' uses the network after a case has been retrieved to decide which actions should be taken to solve the problem. We can therefore say that CBRDM represents a more ``tightly'' integrated method than the other existing methods, although it only performs the retrieve step of the CBR process.