MACHINE LEARNING IN RHEUMATOLOGY

Aram Karalic, Vladimir Pirnat

Abstract

One of the major problems in the development of expert systems is construction of a knowledge base. Instead of the classical methods (interviewing the experts from a certain area), automatic knowledge synthesis can be used to speed up the process. We constructed a knowledge base for rheumatology in the form of decision trees by methodology of learning from examples used in the system ASSISTANT. Knowledge was verified by testing the classification accuracy of the obtained decision trees on new examples (patients). We also compared classification accuracy of induced trees with that of the general practitioners.

Paper.ps