基于随机抽样的模糊粗糙约简

发布时间：2018-01-22 02:21

本文关键词： 模糊粗糙集随机抽样属性约简统计粗糙集　出处：《软件学报》2017年11期 　论文类型：期刊论文

【摘要】：传统的属性约简由于其时间复杂度和空间复杂度过高,几乎无法应用到大规模的数据集中.将随机抽样引入传统的模糊粗糙集中,使得属性约简的效率大幅度提升.首先,在统计下近似的基础上提出一种统计属性约简的定义.这里的约简不是原有意义上的约简,而是保持基于统计下近似定义的统计辨识度不变的属性子集.然后,采用抽样的方法计算统计辨识度的样本估计值,基于此估计值可以对统计属性重要性进行排序,从而可以设计一种快速的适用于大规模数据的序约简算法.由于随机抽样集以及统计近似概念的引入,该算法从时间和空间上均降低了约简的计算复杂度,同时又保持了数据集中信息含量几乎不变.最后,数值实验将基于随机抽样的序约简算法和两种传统的属性约简算法从以下3个方面进行了对比:计算属性约简时间消耗、计算属性约简空间消耗、约简效果.对比实验验证了基于随机抽样的序约简算法在时间与空间上的优势.
[Abstract]:Traditional attribute reduction can hardly be applied to large-scale data sets because of its high complexity in time and space. Random sampling is introduced into the traditional fuzzy rough sets. The efficiency of attribute reduction is greatly improved. First of all, on the basis of statistical approximation, a definition of statistical attribute reduction is proposed. The reduction here is not the original sense of reduction. It is a subset of attributes that keep the statistical identification degree invariant based on the approximate definition under statistics. Then, the sample estimation of statistical identification is calculated by sampling method. Based on this estimate, the importance of statistical attributes can be sorted, and a fast order reduction algorithm for large-scale data can be designed, because of the introduction of random sampling set and statistical approximation concept. The algorithm reduces the computational complexity of the reduction in both time and space, while keeping the information content in the dataset almost unchanged. Finally. Numerical experiments compare the order reduction algorithm based on random sampling with two traditional attribute reduction algorithms from the following three aspects: computing attribute reduction time consumption and computing attribute reduction space consumption. The comparison experiment shows that the order reduction algorithm based on random sampling has advantages in time and space.
【作者单位】：中国人民大学信息学院;中国人民大学数据工程与知识工程教育部重点实验室;中国人民大学环境学院;
【分类号】：TP18
【正文快照】： 随着大数据时代的来临,数据挖掘技术蓬勃发展.近年来,由于不确定性数据比重的不断增大,不确定性数据挖掘越来越受到人们的重视.在不确定数据上进行降维,如基于模糊粗糙集的属性约简,近几年得到广泛关注.但是,现有的模糊粗糙集约简方法由于其基础理论复杂度的桎梏,无法直接应用

【相似文献】