Abstract: This paper proposes a new attribute reduction algorithm for information systems.For that，the paper，firstly constructs the correlation between information systems and relation matrix，studies new and relatively reasonable formulas measuring attribute significance from the viewpoint of relation matrix，and uses the new formulas as heuristic information to design a new attribute reduction algorithm.Compared with the existed attribute reduction algorithm，the algorithm developed in this paper can avoid repeatedly calculating important value of unimportant attributes by gradually deleting unimportant attributes from searching space.Besides，its time complexity has been analyzed in detail，and a real example and experimental results demonstrate both its feasibility and effectiveness.

Key words: rough set theory, information systems, attribute reduction, time complexity

摘要： 提出了一种新的信息系统属性约简算法。为此，首先建立了信息系统与关系矩阵之间的联系；其次，从关系矩阵的角度研究了合理刻画属性重要性的新指标；然后利用新指标作为启发式信息设计了一种新的属性约简算法。与现有算法相比，该算法具有较大的灵活性，它能从搜索空间中逐次删除不重要属性，避免对其重要性的重复计算。此外，对该算法的时间复杂度进行了详细的分析，并通过实例和实验验证它的可行性与有效性。

关键词: 粗糙集理论, 信息系统, 属性约简, 时间复杂性