基于改进模糊决策粗糙集的最小化决策代价特征选择算法  

MINIMUM DECISION COST FEATURE SELECTION ALGORITHM BASED ON IMPROVED FUZZY DECISION-THEORETIC ROUGH SET MODEL

作　　者：王鹏 王玉红 Wang Peng;Wang Yuhong(College of Mathematics and Computer Science,Chifeng University,Chifeng 024000,Inner Mongolia,China)

机构地区：[1]赤峰学院数学与计算机科学学院,内蒙古赤峰024000

出　　处：《计算机应用与软件》2021年第1期284-292,296,共10页Computer Applications and Software

基　　金：高校服务地方基础教育实践研究项目(cfxyfc201853)。

摘　　要：模糊决策粗糙集是决策粗糙集理论在模糊集环境下的重要延伸,然而该模型对含噪声的数据不具有很好的容忍性。为此在传统的模糊相似关系中引入一个限定阈值,提出一种改进的模糊相似关系。在其基础上对原始的模糊决策粗糙集进行重构,提出一种改进的模糊决策粗糙集模型。根据不同的特征选择方式,利用所提出的改进模型设计出两种搜索策略的最小化决策代价特征选择算法。实验分析表明,该算法比传统算法具有更高的优越性。Fuzzy decision-theoretic rough set is an important extension of decision-theoretic rough set theory in the environment of fuzzy sets,but this model does not tolerate noisy data very well.In order to improve this limitation,a limited threshold is introduced into the traditional fuzzy similarity relation,and an improved fuzzy similarity relation is proposed.Then the original fuzzy decision-theoretic rough set is reconstructed on the basis of it,and an improved model of fuzzy decision-theoretic rough set is proposed.According to different feature selection methods,two search strategies are designed to minimize the decision cost feature selection using the proposed improved model.The experimental analysis shows that this algorithm has higher superiority than the traditional algorithm.

关 键 词：决策粗糙集 模糊相似关系 阈值 决策代价 特征选择 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]