FUZZY ROUGH SETS - APPLICATION TO FEATURE-SELECTION

Fuzzy rough sets are a notion introduced as a further extension of the idea of rough sets. Its purpose is fuzzy pattern recognition, where classes are described by fuzzy sets on the feature space. A measure of classification accuracy, which can serve also as a criterion for feature selection, is developed in order to be used in fuzzy recognition problems. It is shown that the fuzzy concepts of Positive, Negative and Boundary region of a set, proposed in the paper, coincide with the original definitions if appropriately applied to ordinary rough classification. Feature selection via fuzzy rough sets is illustrated using real data from medical practice. The problem of evaluating the hypoxic resistance of a patient on the basis of the values of his blood pressure during a barocamera examination was considered. The measurements were evaluated by the fuzzy rough sets criterion. The results obtained are closely related with the expert opinion on the problem.