基于启发式二分策略的属性约简方法

分辨矩阵为属性约简与求核运算提供了一种规范精确的数学模型,通过分辨矩阵模型可方便地获取决策系统全体属性约简.本文基于分辨矩阵运用启发式信息与二分策略扩展结点,设计了一种构造约简树求取全体属性约简的有效算法,该约简树从根结点到叶结点全体路径构成的析取范式与系统分辨函数等价,其极小析取范式所有析取项即为决策系统全体属性约简.该方法适用于满足任意约简准则的分辨矩阵,能够显著地减少搜索空间和保证全体约简求解的完备性,理论分析与实验结果说明了算法的可行性与有效性.

On Method of Attribute Reduction Based on Heuristic Binary Strategy

In the method of discernibility matrix, a kind of normative and precise mathematical model can be provided for finding the core and attribute reduction, which is capable of acquiring all attribute reductions conveniently. An efficient approach based on discernibility matrix is designed for constructing reduction tree by means of heuristic information and binary strategy. In this binary tree, all paths from root to leaf are corresponding to all disjunctors of disjunctive form which is equivalent to discernibility function, and all disjunctors of its extreme minimal disjunctive form are corresponding to all reductions of decision system. The proposed algorithm is applicable to the discernibility matrix satisfying any reduction criterion, and be capable of reducing the searching scope significantly, and finding out all reductions definitely. The theoretical analysis and experimental results show the feasibility and effectiveness of the proposed algorithm.