Increasing Diagnostic Accuracy by Meta Optimization of Fuzzy Rule Bases

In medicine the decision on which test to choose for a given decision problem is a delicate problem. On the one hand a positive test should be a reliable indicator on the presence of a disease, while on the other hand a negative test is required to be an indicator on the absence of a disease. Of course, these two goals are conflicting and a balanced decision according to the current situation is required. Inductive learning methods for (fuzzy) rule bases are, however, typically not capable of optimizing such complex and problem depending goal functions. We therefore present a meta-learning algorithm which selects a subset from a previously generated set of fuzzy rules using bacterial evolutionary algorithms. We also present a study where the proposed method is used to generate a model for predicting the presence/absence of hepatitis, based on laboratory results.