Extending the Measure of Rough Dependency for Fuzzy Classification

In rough-set-based data analysis, the so-called approximation quality is the traditional measure to evaluate the classification success of attributes in terms of a numerical evaluation of the dependency properties generated by these attributes. To deal with practical situations where a fuzzy classification must be approximated by available knowledge expressed in terms of a Pawlak’s approximation space, we introduce in this paper an extension of this measure aimed at providing a numerical characteristic for such situations. Other related coefficients as precision and significance are also discussed correspondingly. A simple example is given to illustrate the proposed notions.