基于k近邻属性重要度和相关系数的属性约简  

Attribute reduction based on k-nearest neighbor attribute importance and correlation coefficient

作　　者：林芷欣 刘遵仁 纪俊 LIN Zhi-xin;LIU Zun-ren;JI Jun(College of Computer Science and Technology,Qingdao University,Qingdao 266071,China)

机构地区：[1]青岛大学计算机科学技术学院,山东青岛266071

出　　处：《计算机工程与设计》2020年第9期2488-2494,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61503208)。

摘　　要：为提高邻域粗糙集属性约简算法的运行效率,降低属性约简算法的复杂度,提出一种基于k近邻属性重要度的约简算法。通过计算某属性下距离样本点最近的k个同类和k个异类样本点的距离得到一种属性重要度的评价指标,融入相关系数方法去除约简属性的冗余信息。通过在多个UCI数据集上进行验证,实验结果表明,与现有算法相比,该算法能更快速判断出数据集中各属性的重要度,有效降低了约简算法的时间复杂度,能更快速得到约简属性,获得较高的分类精度。To improve the efficiency of neighborhood rough set attribute reduction algorithm and reduce the complexity of attri-bute reduction algorithm,a reduction algorithm based on k-neighbor attribute importance was proposed.An evaluation index of attribute importance was obtained by calculating the distance between k similar and k dissimilar sample points closest to sample points under certain attribute,and the correlation coefficient method was adopted to remove redundant information of reduction attribute.Through verification on multiple UCI data sets,the results show that compared with the existing algorithms,the proposed algorithm can judge the importance degree of each attribute in the data set more quickly,effectively reduce the time complexity of the reduction algorithm,get the reduction attribute more quickly,and obtain higher classification accuracy.

关 键 词：邻域粗糙集 属性约简 K近邻 属性重要度 相关系数 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]