基于多粒度模糊邻域熵的在线流组特征选择  

Online group streaming feature selection based onmulti-granularity fuzzy neighborhood entropy

作　　者：韩子钦 徐久成 章磊 周长顺 许诗卉 HAN Zi-qin;XU Jiu-cheng;ZHANG Lei;ZHOU Chang-shun;XU Shi-hui(College of Computer and Information Engineering,Henan Normal University,Xinxiang 453007,China;Engineering Lab of Intelligence Business and Internet of Things of Henan Province,Henan Normal University,Xinxiang 453007,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007 [2]河南师范大学智慧商务与物联网技术河南省工程实验室,河南新乡453007

出　　处：《计算机工程与设计》2025年第1期214-222,共9页Computer Engineering and Design

基　　金：国家自然科学基金项目(61976082、62076089、62002103)。

摘　　要：针对传统在线流组特征选择方法无法处理异常或者缺失的不完备混合数据,导致特征选择效果不佳的问题,在不完备系统中提出一种基于多粒度模糊邻域熵的在线流组特征选择算法。考虑不完备混合数据中的不确定信息,将决策自信息与模糊邻域熵相结合,从代数和信息视角提出多粒度模糊邻域熵;提出在线流组内、组间粒选度,根据模糊邻域对比度对特征组进行冗余分析。在8个公共数据集上进行实验对比分析,所提算法在处理不完备混合数据时能有效消除冗余特征,提高数据的分类精度。Aiming at the problem that traditional online stream group feature selection methods cannot deal with abnormal or missing incomplete mixed data,which leads to poor feature selection,an online group streaming feature selection algorithm based on multi-granularity fuzzy neighborhood entropy was proposed in incomplete systems.Considering the uncertain information in incomplete mixed data,the decision self-information was combined with the fuzzy neighborhood entropy,and the multi-granularity fuzzy neighborhood entropy was proposed from both algebraic and informational perspectives.The intra-granularity and inter-granularity selection degree was proposed in online group streaming,and the redundancy analysis was performed for the feature groups based on the fuzzy neighborhood contrast.Experimental comparative analyses were carried out on eight public datasets,the proposed algorithm can dramatically remove redundant features and raise the classification accuracy of data when dealing with incomplete mixed data.

关 键 词：流特征选择 流组 自信息 模糊邻域粗糙集 不完备决策系统 模糊邻域熵 重合度 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]