A probabilistic inductive learning approach to the acquisition of knowledge in medical expert systems

Abstract

An inductive knowledge acquisition method based on the probabilistic inference technique is presented. The proposed system can be applied to generate decision rules automatically for certain medical expert systems. Given a patient database containing historical diagnosis and prognosis information, the method is capable of detecting the inherent probabilistic patterns in the data. Classification knowledge can be synthesized in the form of explicit production rules with associated probabilistic weight of evidence based on the patterns detected. With these rules, new patient cases can be quickly and accurately classified. Using real-world medical data, it is shown that the proposed method performs better in terms of classification accuracy and computational efficiency than some of the major existing methods.<>