基于不完备区间值信息系统的决策粗糙集

Decision-theoretic rough set based on incomplete interval-valued information system

在不完备区间值信息系统中,提出一种基于极大相容类的决策粗糙集模型。针对不完备区间值信息系统中属性相似度存在的缺陷,对属性相似度进行改进。在不完备区间值信息系统中,由于容差关系下建立粗糙集模型存在冗余度高、分类精度低的问题,采用极大相容类代替等价类,结合贝叶斯最小风险决策原则,建立决策粗糙集模型。经证明,基于极大相容类建立粗糙集模型可有效提高分类精度。最后,基于正域分布不变的原则提出基于区分矩阵的属性约简算法并将该算法应用于实例。

In incomplete interval-valued information system, this paper proposed the decision-theoretic rough set based on maximal consistent class. Considering the insufficient about attribute similarity in incomplete interval-valued information system, it provided the improved attribute similarity. Then, in order to solve the model＇ s high redundancy and low classification accuracy in the information system, this paper replaced equivalence class with the maximal consistent class and set up the decision-theoretic rough set model combined with Bayesian smallest risk theory. And it proved that set up the model based on maximal consistent class can improve the classification accuracy. Finally, it proposed the attribute reduction algorithm based on indiscernibility matrix and remains distribution of positive region and applied it to a case.