基于粗糙集的约简算法及规则融合方法的研究

Study on Rough Set Theory-Based Reduction Algorithm and Method of Rule Fusion

从不同的角度分析了属性约简的两种重要方法:区分矩阵法和基于属性重要性。根据数据集的实际情况提出了一种基于粗糙集的区分矩阵和属性重要性相结合的启发式算法,并获得了属性约简集。在约简集的基础上分析了静态决策推理规则及算法。在相容决策系统中利用集合向量包含度构造了规则融合的方法,从而得到动态条件规则的极大近似决策值。在知识满足分类质量要求的前提下,根据规则融合方法,对任意给定的样本知识可以判别知识的实际归属类。

Discernibility matrix and the significance of attributes were analyzed from different aspects, both of them were important methods of attribute reduction. A heuristic algorithm on rough set based on coalescent of discernibility matrix and the significance of attributes was raised. According to the algorithm a relative reduction set was formed. On the basis of the reduction set, the static decisioa rule reasoned and algorithm were analyzed. Finally the methood of rule fusion was constructed by inclusion degree of set vector in the consistent decision system. The maximum approximate decision values were acquired from dynamic conditional rules. Under the condition of knowledge satisfied with classification quality, aceording to methods of rule fusion,its actual adseription class was found for any known sample knowledge.