To exemplify an integrated CBR-DM systems, we used Bayesian Knowledge Discoverer to learn the network structure of a database, and implemented a simple program (CBRDM) described in the previous section. To compare results, we used another CBR tool, KATE. Bayesian Knowledge Discoverer was chosen because it produces graphs in an easy to handle format, Bayesian networks interchange format. It is a research prototype that is freely available, and is relatively fast. KATE was selected because it is easy to use, and because the manufacturer, Acknosoft, is one of the NOEMIE partners. Creek [1], which is a knowledge-intensive CBR system developed at NTNU might also have been used, but KATE was chosen because of easy integration with Bayesian networks in C++. A list of other existing tools can be found in Appendix F. We also used Microsoft Belief Networks for manual inspection of the network produced by BKD. We give a brief description of each of the tools: