Automated discovery of positive and negative knowledge in clinical databases

The rule-induction methods are categorized into deteministic and probabilistic rules. The deterministic rules are described as if-then rules, viewed as prepositions while the probabilistic rules are if-then rules with probabilistic information. The characteristics of medical reasoning, introduced as positive and negative rules, are established as the models of medical reasoning. Two algorithms for induction of positive and negative rules are defined as search procedures by using accuracy and coverage as evaluation indices. The evaluation of proposed method on medical databases show that induced rules correctly represented experts' knowledge and also uncovered several interesting patterns.