关键词: 粗糙集, 粒计算, 属性约简, 属性重要性

Abstract: Granular computing is a new intelligent computing theory and method based on issue partition concepts.Inconsistent decision table is an important aspect in rough set theory.Using the equivalent relation in rough set to construct particle，this paper offers decompounds method of particle，furthermore gives the heuristic information in attribute reduction algorithm based on attribute importance.At last，experimental result shows that the new algorithm is not only very efficient but also can treat with large decision table.

Key words: rough set, granular computing, attribute reduction, attribute importance