不一致邻域粗糙集的不确定性度量和属性约简

Uncertainty Measure and Attribute Reduction in Inconsistent Neighborhood Rough Set

不确定性度量和属性约简是邻域粗糙集模型中重要的研究内容.针对目前已有的粗糙集不确定性度量方法难以应用于邻域粗糙集中,同时考虑到现有属性约简算法中很少考虑条件属性之间的关系也会影响约简结果和分类精度.首先分析了不一致邻域粗糙集的相关性质,然后提出了邻域条件熵的不确定性度量方法用来评价约简属性的质量,分析证明了相关的性质定理,接着引入统计学中秩相关系数的概念,通过计算条件属性之间的相关系数来剔除冗余属性,构造了基于相关系数的不一致邻域粗糙集属性约简算法(RNRS).最后在UCI数据集上与现有算法进行了比较分析,实验结果表明,本文的算法可以获取较少的属性特征和较高的分类精度.

Uncertainty measure and attribute reduction are the important research contents in neighborhood rough set model.In this paper,it is difficult to apply the rough set uncertainty measure method to the neighborhood rough set.Considering that the relationship between the conditional attributes is rarely considered in the existing attribute reduction algorithm,the reduction result and classification accuracy are also affected.we first analyze the correlation properties of the inconsistent neighborhood rough sets,and then propose the uncertainty measure method of neighborhood entropy to evaluate the quality of the reduced attributes.The related property theorems are proved and then the rank correlation The concept of coefficient is used to eliminate the redundant attributes by calculating the correlation coefficient between conditional attributes,and the attribute reduction algorithm of inconsistent neighborhood rough set based on correlation coefficient（RNRS） is constructed.Finally,the UCI data sets are compared with the existing algorithms.The experimental results show that the algorithm can obtain fewer attribute characteristics and high classification accuracy.