基于条件信息熵的多值信息系统的属性约简

从信息论的角度提出了多值决策信息系统的信息熵和条件信息熵的概念,在此基础上提出了一种属性约简算法,以一种新的形式的属性重要性为启发式信息,减小了知识约简过程中的搜索空间,实验分析表明,该算法能得到多值信息系统的最小约简。

Attribute Reduction of Multivalued Information System Based on Conditional Information Entropy

In this paper,authors discuss the information entropy and conditional information entropy of multivalued information system from the viewpoint of information theory,then,an attribute reduction algorithm is proposed based on a novel significance of attributes,which lessens search space in the process of reduction.Simulation results show that the valid algorithms can find the minimal reduction in most cases.