关键词: 粗糙集, 优势关系, 知识粒度, 近似精度, 近似约简

Abstract: A concept of knowledge granularity in rough sets based on dominance relation is introduced，and conclusion is demonstrated that knowledge granularity declines as the enhancement of uncertain extent of knowledge.On this basis，roughness and accuracy are defined in the information systems based on dominance relation，approximate accuracy in inconsistent objective information systems based on dominance relation.Conclusion is proved that accuracy and approximate accuracy can measure the approximate property of the importance.Furthermore a heuristic reduction algorithm with approximate accuracy is heuristic information raised inconsistent objective information systems based on dominance relation，and the time complexity of the algorithm is analysed as follows.Finally，an example illustrates this algorithm is practical and effective.

Key words: rough sets, dominance relation, knowledge granularity, approximate accuracy, approximation reduction