Decision Rule Generation Based on Similarity Relation*

The indiscernibility relation is the mathematical basis for the classical rough set theory. It is natural to extend the indiscernibility relation originally used for the definition of the rough approximation when the data describing objects is imprecise or when small differences are meaningless in the context of the study. This situation can be modeled using a binary relation that represents a certain form of similarity. Based on the concepts of lower and upper approximations using the similarity relation, the nonsimilarity matrix and the nonsimilarity function are presented to induce the minimal decision rules. An example is illustrated to demonstrate the application of this new approach