基于信息熵和差别矩阵的属性约简方法

Attribute reduction method based on information entropy and discernibility matrix

针对无决策属性的信息系统,定义了无决策属性的差别矩阵,利用商空间理论下信息熵的计算方法,提出了一种“空不变”无决策属性约简方法。该方法从属性约简的本质出发,将属性造成等价类发生改变的可能性数值化,进而实现属性约简。通过实例验证了方法的有效性,并通过与已有方法的对比和实验结果分析,进一步证实了该方法能更准确地识别出关键属性,完成属性约简。

For the information system without decision attributes,the difference matrix without decision attributes is defined,and a“empty invariant”reduction method without decision attributes is proposed by using the calculation method of information entropy under the quotient space theory.Starting from the essence of attribute reduction,the probability of attribute causing the change of the equivalent class is quantified,and then the attribute reduction is realized.The effectiveness of the method is verified by examples,and the comparison with the existing methods and the analysis of experimental results further confirm that the method can more accurately identify the key attributes and complete the attribute reduction.