ALL libraries (COBIB.SI union bibliographic/catalogue database)
  • Learning rules for early diagnosis of rheumatic diseases
    Lavrač, Nada ...
    The paper describes the appliction of our inductive learning system LINUS to the problem of learning rules for early diagnosis of rheumatic diseases. Diagnostic knowledge for this domain was ... automatically constructed in earlier experiments with ASSISTANT, a program which inductively learns attribute descriptions in the form of decision trees. The attribute-value learning approach is limited by the restricted representation formalism and the inability to use background knowledge. LINUS uses attribute-value learning algorithms, but incorporates them into a more powerfull logical framework, where background knowledge can be used in induced concept descriptions. The results obtained with LINUS incorporating ASSISTANT, using background knowledge provided by a medical specialist, are presented. An analysis is given of how different noise-handling mechanism and the use of background knowledge affect classification accuracy and complexity of diagnostic rules.
    Type of material - conference contribution
    Publish date - 1991
    Language - english
    COBISS.SI-ID - 2699798