伪标签邻域粗糙集下的属性约简加速策略  

Acceleration strategy for attribute reduction based on pseudo-label neighborhood rough set

作　　者：饶先胜 宋晶晶[1,2] 杨习贝 于化龙[1] 王平心[3] RAO Xian-sheng;SONG Jing-jing;YANG Xi-bei;YU Hua-long;WANG Ping-xin(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212003,China;Fujian Province Universities Key Laboratory of Data Science and Intelligence Application,Minnan Normal University,Zhangzhou 363000,China;School of Science,Jiangsu University of Science and Technology,Zhenjiang 212003,China)

机构地区：[1]江苏科技大学计算机学院,江苏镇江212003 [2]闽南师范大学数据科学与智能应用福建省高校重点实验室,福建漳州363000 [3]江苏科技大学理学院,江苏镇江212003

出　　处：《计算机工程与设计》2020年第11期3087-3093,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61572242、61906078);数据科学与智能应用福建省高校重点实验室开放课题基金项目(D1901)。

摘　　要：为降低伪标签邻域粗糙集中求解一组半径下约简的时间消耗,在基于贪心策略的启发式搜索基础上,通过减少属性约简过程中属性的遍历规模,设计一种约简求解的加速策略。在求解当前半径下的约简时,其启发式搜索过程是在前一个半径所求得约简结果基础上,继续选择重要度最大的候选属性加入当前约简中。在8个UCI数据集上的实验结果表明,相较于使用启发式算法求解一组半径下的约简,所提加速策略在不降低约简性能的同时,能有效减少求解一组半径下约简的时间消耗。该方法为快速求解伪标签邻域粗糙集的约简提供了技术支撑。To reduce the time consumption of finding reduction over a set of radii in pseudo-label neighborhood rough set,an acceleration strategy was proposed.Such acceleration strategy was designed by reducing the traversal scale of attribute in the process of greedy strategy based heuristic searching.The process of finding current radius based reduction was realized by conti-nuing to select the most significant attribute from the remained candidate attributes,and then adding it into the current reduction.Experimental results over 8 UCI data sets show that the proposed acceleration strategy can effectively decrease the elapsed time of finding reduction over a group of radii without reducing the classification performance.This method provides technical support for quickly finding reduction in pseudo-label neighborhood rough set.

关 键 词：加速策略 属性约简 启发式算法 邻域粗糙集 伪标签 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]