基于分布约简的差别矩阵方法研究

Algorithm of Discernibility Matrix Based on Distribution Reduction

在不相容信息系统中,代数观点与信息论观点下的属性约简是不等价的。阐述了分布约简与信息论观点下约简的等价性,分析了代数观点下差别矩阵的缺陷,提出了基于分布约简的差别矩阵方法。该方法不仅考虑所比较对象是否不一致,并且深入分析其不一致性的程度及所属条件等价类在决策类中的分布比例,得到信息系统在分布约简概念下的约简。理论分析与仿真实例表明了该算法的可行性和有效性。

In inconsistent information systems, the reduction in algebra view are not equivalent to that in information view. The conclusion is expounded that distribution reduction is view, the problem of discernibility matrix in algebra view is analyzed, based on distribution is proposed in this paper. This algorithm considers ent, analyses in detail the classes in decision classes, sis and simulation instance degree of inconsistency and the distributing equivalent to the reduction in information and an algorithm of discernibility matrix whether the objects compared are consistproportion of their conditional equivalent and the reduction in distribution reduction view is acquired finally. The theoretic analyshows that this algorithm is feasible and effective in practice.