Abstract
By using mathematical logic and the fuzzy technique, we present an approach to the computation of optimal feature sets for diagnostic tests in microbiology. These sets of physiological and biochemical tests are suitable for differentiating all organisms belonging to a given group. Membership functions are defined which are able to reflect the time dependence of the test reaction states. The result is a suboptimal solution characterized by short feature sets of high diagnostic value. An application to Rhodotorula species is shown.