Abstract: In intuitionistic fuzzy relations, between objects have more superior relationship evaluation effect through the description of membership degree and non-membership degree. In order to achieve better measurement for information system. Firstly, the intuitionistic fuzzy granulation for information system based on intuitionistic fuzzy relations is introduced in this paper. And then, the concepts of entropy are respectively defined according to membership degree and non-membership degree in granulation results, in addition, combining them as the information entropy of the information system under the intuitionistic fuzzy relation. Finally, an attribute reduction algorithm is constructed according to the information entropy. Experimental results show that the proposed algorithm has better attribute reduction performance.