基于不完备决策信息系统的知识约简算法

Knowledge Reduction Algorithm Based on Incomplete Decision Information System

针对复杂领域中的决策信息系统往往存在数据缺失现象,差别矩阵作为决策信息系统知识约简的有效方法之一,但传统的差别矩阵方法并不适用于不完备决策信息系统,为此构造了基于不完备决策信息系统的差别矩阵,并简化了差别矩阵中的属性元素,有效地删除了差别矩阵中的冗余元素。然后结合不完备决策信息系统的核属性,利用启发式方法设计了一种基于不完备决策信息系统的快速知识约简算法,最后通过实验比较验证了算法的有效性。

The data in many decision systems of actual complex fields are usually incomplete, while discernibility matrix is one of the most significant methods of feature reduction in decision systems; traditional discernibility matrix is inefficient way to extract knowledge reduction from incomplete decision information system. To solve such problem, discernibility matrix based on tolerance is constructed based on incomplete decision information system, and the absorption is used to simplify elements in discernibility matrix, which delete many redundant elements. What’s more, core attribute is used in decision information system. On this condition, quick knowledge reduction algorithm is designed based on incomplete decision information system by heuristic method. The experiential results show that the proposed algorithm is efficient.