Information Refinement and Revision for Medical Expert System -- Automated Extraction of Hierarchical Rules from Clinical Data

Since real-world decision making include several decision steps, real-world decision making agents have several sophisticated diagnostic reasoning mechanisms. Thus, if we want to update these reasoning steps, we have to extract rules for each step from real-world datasets. However, one of the most important problems on rule induction methods is that they aim at induction of simple rules and cannot extract rules that plausibly represent experts decision processes, which makes rule induction methods not applicable to the maintenance of real-world decision making agents. In this paper, the characteristics of experts rules are closely examined and a new approach to extract plausible rules is introduced, which consists of the following three procedures. First, the characterization of decision attributes (given classes) is extracted from databases and the classes are classified into several groups with respect to the characterization. Then, two kinds of sub-rules, characterization rules for each group and discrimination rules for each class in the group are induced. Finally, those two parts are integrated into one rule for each decision attribute. The proposed method was evaluated on medical databases, the experimental results of which show that induced rules correctly represent experts decision