差别矩阵为粗糙集属性约简提供了很好的思路,但差别矩阵中存在冗余的重复和父集元素.为了消除这些冗余元素,提出一棵有序树:差别信息树,该树能消除差别矩阵中的重复元素,同时在大多数情况下也能完全消除父集元素,实现对差别矩阵中非空元素的压缩存储.为了验证差别信息树的有效性,提出一种属性约简完备算法,并使该算法的时间复杂度降为O(∣C∣∣U∣2). Difference matrices provide a good idea for rough set attribute reduction, but redundant matrices and parent elements exist in the discernibility matrix.In order to eliminate these redundant elements, an ordered tree is proposed: a difference information tree, In order to verify the effectiveness of the difference information tree, a new method of attribute reduction is proposed, which can eliminate the repeated elements in the difference matrix and eliminate the parent element completely in most cases. Algorithm, and reduce the time complexity of the algorithm to O (|C||U|2).