Mining diagnostic rules from clinical databases using rough sets and medical diagnostic model

Since rule induction methods generate rules whose lengths are the shortest for discrimination between given classes, they tend to generate rules too short for medical experts. Thus, these rules are difficult for the experts to interpret from the viewpoint of domain knowledge. In this paper, the characteristics of experts' rules are closely examined and a new approach to generate diagnostic rules is introduced. The proposed method focuses on the hierarchical structure of differential diagnosis and 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 generalized groups with respect to the characterization. Then, two kinds of sub-rules, classification rules for each generalized group and rules for each class within each group are induced. Finally, those two parts are integrated into one rule for each decision attribute. The proposed method was evaluated on a medical database, the experimental results of which show that induced rules correctly represent experts' decision processes.