Discovery of Positive and Negative Knowledge in Medical Databases Using Rough Sets

One of the most important problems on rule induction methods is that extracted rules partially represent information on experts’ decision processes, which makes rule interpretation by domain experts difficult. In order to solve this problem, the characteristics of medical reasoning is discussed, and positive and negative rules are introduced which model medical experts’ rules. Then, for induction of positive and negative rules, two search algorithms are provided. The proposed rule induction method was evaluated on medical databases, the experimental results of which show that induced rules correctly represented experts’ knowledge and several interesting patterns were discovered.