机构地区: 佛山科学技术学院理学院
出 处: 《华中科技大学学报(自然科学版)》 2012年第4期8-12,共5页
摘 要: 针对决策表,首先从理论上证明了相对知识量与新条件信息熵的等价性,从而保证可以从相对知识量的角度计算Hu差别矩阵属性约简.为快速计算出决策表属性的相对知识量,给出一种相对知识量的递归计算公式,从而得到新条件信息熵的递归计算公式.以此为基础给出一种属性重要性定义并设计出一种计算Hu差别矩阵属性约简的启发式算法.算例和实验结果验证该算法是高效的. For a decision table, it was firstly proved that the relative knowledge quantity was equivalent to new conditional information entropy and it was guaranteed that the attribute reduction based on Huts discernibility matrix could be calculated by relative knowledge quantity. To get the relative knowledge quantity more quickly, a recursive formula was proposed and the corresponding one of conditional information entropy was gotten. Then a signification of attribute was defined and the heuristic algorithm was designed for the attribute reduction based on Huts discernibility matrix. Numerical ex- amples and experimental results illustrate the efficientibility.
领 域: [自动化与计算机技术] [自动化与计算机技术]