中文摘要：

知识约简可以保持决策系统中的分类特征不变,是粗糙集理论的重要研究内容之一。分布约简保持约简前后决策系统中各规则的置信度不发生改变。为了给区间值决策系统的论域分类提供合理的度量标准,引入了区间值相似率。通过将Pawlak决策系统中的等价关系扩展到区间值决策系统中的相容关系,提出了区间值决策系统的分布约简目标。针对该目标给出了相应差别矩阵的计算方法,并与现有区间值决策系统的广义决策约简计算方法进行了分析比较。最后,通过人工数据集的实验验证了相关结论的有效性。

英文摘要：

Different classification features in decision systems can be kept by knowledge reduction which is one of the hottest issues in rough set theory. The confidence level is unchanged because of distribution reduction in decision sys- tems. For providing the measure criterion for universe classification in interval valued decision systems, the similarity coefficient was introduced in this paper. To extend the equivalence relation in Pawlak decision systems to the tolerance relation in interval-valued decision systems, we proposed the concept of distribution reduction in inconsistent interval valued decision systems. Aiming at the proposed concept, we provided the computational method of corresponding dis- cernibility matrix. We also discussed the relation of distribution reduction and generalized decision reduction in interval valued decision systems. Finally, experiments show that the novel method is effective.