一种新的决策粗糙集最小化决策代价属性约简算法

A new decision-theoretic rough set model and minimum decision cost attribute reduction algorithm

实际应用中存在着大量的数值型数据,然而传统的决策粗糙集只能够处理符号型数据,为了改善这一局限性,本文构造出一种模糊邻域决策粗糙集模型,并提出一种最小化决策代价的属性约简算法.文中首先将将模糊粗糙集和邻域粗糙集融入决策粗糙集中,提出了模糊邻域决策粗糙集,使得该模型同时具有模糊粗糙集和邻域粗糙集处理数值型数据的优点;然后基于该模型,给出一种决策代价定义,并提出相应的最小化决策代价属性约简算法;最后通过实验分析表明所提出的算法具有较好的代价敏感属性约简性能.

There are a lot of numerical data in practical application,but traditional decision-theoretic rough sets can only deal with symbolic data.In order to improve this limitation,this paper constructs a fuzzy neighborhood decision-theoretic rough set model and proposes a minimum decision cost attribute reduction algorithm.Firstly,the fuzzy rough set and neighborhood rough set are integrated into decision-theoretic rough set,and the fuzzy neighborhood decision-theoretic rough set is proposed,which makes the model have the advantages of both fuzzy rough set and neighborhood rough set in dealing with numerical data.Then,based on the model,a definition of decision cost is given and the corresponding minimum decision cost attribute reduction is proposed.Finally,the experimental results show that the proposed algorithm has better cost-sensitive attribute reduction performance.