Extracting diagnostic taxonomy using rough sets

One of the most important features of medical expert reasoning is that each reasoning rule is composed of several diagnostic steps, usually hierarchical differential diagnosis. In this paper, the characteristics of experts´ rules are closely examined from the viewpoint of hierarchical decision steps. Then, extraction of diagnostic taxonomy from medical datasets is introduced, which consists of the following three procedures. First, the characterization set of each decision attribute (a given class) is extracted from databases. Then, similarities between characterization sets are calculated. Finally, the concept hierarchy for given classes is generated from the similarity values.