摘要
提出了一种基于论域离散度的适应度函数,在前向贪心搜索策略下,以该适应度函数评估条件属性的重要性,进而求取邻域粗糙集的约简。该算法与3个比较流行的属性约简算法进行对比实验,在12组UCI数据集上进行验证。实验结果表明,与另外3种算法相比,在不降低分类效果的情况下,本文算法在时间消耗和稳定性上具有较为明显的优势。
A fitness function based on the dispersion of the universe was proposed.Under the forward greedy searching strategy,the fitness function was employed to evaluate the importance of conditional attributes and then used to derive the reduct of the neighborhood rough set.The proposed algorithm was compared with three state-of-the-art attribute reduction algorithms and validated on 12 UCI datasets.Experimental results showed that compared with the other three algorithms,the proposed algorithm exhibited better in time consumption and stability without reducing the classification effect.
作者
刘长顺
刘炎
宋晶晶
徐泰华
LIU Changshun;LIU Yan;SONG Jingjing;XU Taihua(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212100,Jiangsu,China;Key Laboratory of Oceanographic Big Data Mining&Application of Zhejiang Province,Zhejiang Ocean University,Zhoushan 316022,Zhejiang,China)
出处
《山东大学学报（理学版）》
CAS
CSCD
北大核心
2023年第5期26-35,52,共11页
Journal of Shandong University(Natural Science)
基金
国家自然科学基金资助项目(61906078,62006099,62076111,62006128)
江苏省高等学校自然科学基金资助项目(20KJB520010)
浙江省海洋大数据挖掘与应用重点实验室开放课题(OBDMA202104)。
关键词
属性约简
条件属性
适应度函数
邻域粗糙集
attribute reduction
conditional attribute
fitness function
neighborhood rough set
作者简介
刘长顺(1992-),男,硕士研究生,研究方向为粒计算、粗糙集等.E-mail:lcs1185156136@163.com;通信作者:宋晶晶(1990-),女,博士,副教授,研究方向为粒计算、粗糙集等.E-mail:songjingjing108@163.com。
