基于模糊-粗糙模型的逼近精度分类规则提取策略

Extracting strategy of impend precision classification rule based on fuzzy-rough set

为了处理具有连续属性的决策系统,利用模糊理论与粗糙理论在处理不确定性问题方面的差异性,提出一种基于模糊-粗糙模型的逼近精度分类规则提取策略.首先利用模糊π函数对决策系统中的连续属性构造三个模糊参数进行模糊化,从而确定条件属性的模糊区域;再根据模糊相似关系构造模糊相似矩阵,然后基于模糊等价类划分的概念,提出了利用逼近精度近似度量的数据挖掘方法进行属性约减,最后应用实例说明如何在决策表中发现分类规则.实验结果表明此方法挖掘出的规则简练且合理可靠.

In order to dealing with the decision system with successive attributes, an abstracting strategy of impend precision classification rule is presented based on the difference between fuzzy and rough set theories. Firstly, the fuzzy region of conditional attribute can be determined while the successive attributes in decision system are fuzzied through constructed with three fuzzy parameters by fuzzy function π and then, relying on fuzzy analogy relation to construct fuzzy analogy matrix and basing on the conception of fuzzy equating classification, a new attributes reduction method of data mining based on impend precision approximation measurement is expounded. Finally, an example on how to discover classification rules in decision table is applied and the result of experimentation shows that the rules are not only simple but also rational.