基于邻域粗糙集的连续值分布式数据属性约简

发布时间：2019-03-09 13:48

【摘要】：为了去除系统中的冗余属性,保持系统的分类能力,研究了连续值分布式数据的属性约简。给出了连续值分布式决策信息系统中邻域粗糙集的定义,讨论了分布式连续值决策信息系统中正域计算的可分解性。以保持分布式决策信息系统的正域不变为前提,探讨了分布式决策信息系统中属性的可约性,提出了分布式连续值决策信息系统的属性约简算法。为了验证该算法的有效性,在7份数据集上进行了3组实验。实验使用提出的算法对分布式数据进行属性约简,进而采用加权集成的方式进行分类测试。实验结果表明,该算法能够有效去除连续值分布式数据中的冗余属性,使得约简后的连续值分布式数据的集成分类能力与约简前相差不大,甚至更高。
[Abstract]:In order to remove redundant attributes in the system and keep the classification ability of the system, the attribute reduction of continuous value distributed data is studied. The definition of neighborhood rough set in continuous-valued distributed decision-making information system is given, and the decomposability of positive-domain computation in distributed continuous-valued decision-making information system is discussed. In order to keep the positive domain of distributed decision-making information system unchanged, the reducibility of attributes in distributed decision-making information system is discussed, and the attribute reduction algorithm of distributed continuous-value decision-making information system is proposed. In order to verify the effectiveness of the algorithm, three sets of experiments were carried out on 7 datasets. The proposed algorithm is used to reduce the attributes of the distributed data, and then the weighted integration method is used to classify and test the data. The experimental results show that the proposed algorithm can effectively remove redundant attributes in continuous value distributed data and make the integrated classification ability of the reduced continuous value distributed data not much different or even higher than that before reduction.
【作者单位】：重庆邮电大学计算智能重庆市重点实验室;
【基金】：国家自然科学基金(61472056,61379114) 教育部人文社科规划项目(15XJA630003) 重庆市教委科学技术研究项目(KJ1500416) 重庆市基础科学与前沿技术研究(cstc2017jcyj AX0406)~~
【分类号】：TP18

【参考文献】