A multigranulation fuzzy rough approach to multisource information systems

Abstract

Multigranulation rough set is one class of the important models in rough set community. However, both pessimistic and optimistic rough sets have disadvantages in describing target concept. In this paper, a novel model, called weighted multigranulation fuzzy decision rough sets, is proposed. Gaussian kernel is used to compute the similarity between objects, which induces a fuzzy equivalence relation. We employ the relation to fuzzily partition the universe and then obtain multiple fuzzy granulations from multisource fuzzy information system. Moreover, some interesting properties of the proposed model are discussed. A comparative study between the proposed multigranulation model and Sun’s multigranulation rough set model is carried out. An example is employed to illustrate the effectiveness of the proposed method, which may provide an effective approach for multisource data analysis in real applications.