相似关系下变精度粗糙集的阈值选取分析  

Analysis of Threshold Value in Variable Precision Rough Sets Based on Similarity Relation

作　　者：谭旭[1,2] 程贲[2] 陈英武[2] 

机构地区：[1]深圳信息职业技术学院计算机应用系,深圳518029 [2]国防科学技术大学信息系统与管理学院,长沙410073

出　　处：《系统仿真学报》2010年第2期403-406,共4页Journal of System Simulation

摘　　要：针对相似关系下的变精度粗糙集模型,讨论了该模型下的阈值选取问题。阈值选取包括相似阈值和分类阈值的选取。通过粗糙熵的计算,给出了最优相似阈值的选取方法。对于分类阈值的选取,提出先将模型的边界域进行细划分,分析各边界域与分类阈值之间的影响关系;之后确立基于边界域的近似分类的准则,通过基于近似分类准则的优化计算,达到对分类阈值区间的合理选取。文末的仿真实例验证了该阈值选取方法的可行性和优越性。Aiming at variable precision rough sets model based on similarity relation, the threshold selection problem was discussed. The threshold selection includes similarity threshold and classification threshold selection. By iteratively computing the defined rough entropy, the optimal similarity threshold could be obtained. As to classification threshold, influence relationships between the classification threshold and the related boundary regions were firstly analyzed after partitioning the boundary region of VPRS model in detail. Then, some approximation classification criterions were acquired based on subdivided boundary regions. Lastly, through optimization computing with the given criterions, the best scope of classification threshold value was obtained. At the last part, example of computing and analyzing with the proposed threshold selection method shows its effectiveness and advantage.

关 键 词：相似关系 变精度粗糙集 阈值 粗糙熵 边界域 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]