To deal with the complex and redundant problem of the evaluation index in the aquatic products safety assessment system, a new attribute reduction algorithm with hybrid information granularity was proposed accordance with attribute unimportance. The traditional attribute reduction algorithms based on rough set usually consider a single factor at present, and its executive efficiency of algorithms is not very ideal. This proposed method in the paper firstly analyzed the feasibility of building attribute insignificance architecture based on rough set and fuse the concepts of positive region and negative region to define the formula of attribute insignificance. Then, it only calculates the attributes unimportance of the assessment system which except the kernel attributes in order to reduce the search space. Finally, through the experimental analysis and validation, it shows that the algorithm is effective and efficient and can be applied to reduce the assessment index in the aquatic products safety assessment system.