Rough set theory provides a useful tool for describing uncertain concepts. The description of a given concept constructed based on rough regions can be used to improve the quality of classification. Processing large data using rough set methods requires efficient implementations as well as alternative approaches to speed up computations. This paper proposes a representative-based approach for rough region-based classification. Positive, boundary, and negative regions are replaced with their representatives sets that preserve information needed for generating classification rules. For data divisible into a relatively low number of equivalence classes representatives sets are considerably smaller than the whole regions. Using a small representation of regions significantly speeds up the process of rule generation.
展开▼
