Abstract: Granular computing is a theory dealing with uncertain data,including rough set,fuzzy set,quotient space,computing with words,etc.At present,the granulation of data and granular computing are mainly related to the set ope-rations.As we know,these set operations are inefficient,resulting in restrict the applications of granular computing algorithms.Therefore,we proposed a binary granular computing model,which has the three layer structure including granule,granule swarm and granule library.We defined binary granules and granule operations,which can transform the set operations into the binary number calculations.Furthermore,we proposed a distance metric of two binary granules,which represents the distance of the set of equivalence classes,and discussed some properties of the granule distance.The binary granular computing model defines the concept of binary granule swarm distance,gives the calculation method of binary granule swarm distance,and puts forward the method of attribute reduction based on binary granule swarm distance.We proved the equivalence of our proposed reduction method and the classical Pawlak reduction method.We presented two kinds of reduction algorithm,which use the binary granule swarm distance as the heuristic information.

Key words: Granular computing,Rough sets,Binary granules,Granule swarm distance