基于二元信道互信息的属性约简方法及其应用

由于知识的粗糙性和信息熵的密切关系,基于互信息的属性约简一直以来都是粗糙集研究的热点,但现有方法都不同程度地依赖于属性值的分布情况,从而影响最终约简结果.为解决此问题,从信息论的角度对属性进行细分,定义了一种新的基于二元信道互信息的属性重要度度量方法,以此为启发式信息构造新的约简算法,实验证明了该算法的有效性;将其应用于重要气象因子的提取,获得了满意的效果.

Attribute Reduction Method Based on Mutual Information of Binary Channel and its Application

As the roughness of the knowledge and entropy has close relationship,attribute reduction based on mutual information is one of the research hotspots.However,the existing methods were different levels dependent on the distribution of attribute values,thus affecting the ultimate reduction results.To solve the problem,this paper defined a new attribute importance measure method based on mutual information of binary channel,then used it as the heuristic information in attribute reduction algorithm.It is testifi...