Knowledge Discovery in Clinical Databases: An Experiment with Rule Induction and Statistics

The main difference between empirical learning methods and KDD methods is that the latter approaches support discovery of knowledge in databases whereas the former ones focus on extraction of accurate knowledge from databases. Therefore, for application of KDD methods, domain experts’ interpretation of induced results is crucial. However, conventional approaches do not focus on this issue clearly. In this paper, several KDD methods are compared by using a common database and the induced results are interpreted by a medical expert, which enables us to characterize KDD methods more concretely and to show the importance of interaction between KDD researchers and domain experts.