基于正区域约简的决策表差别矩阵方法研究

利用差别矩阵比较容易求出决策表的所有属性约简.分析了几种决策表差别矩阵的生成特点,提出一种改进的差别矩阵,差别矩阵的生成不需要计算划分U/C或正区域,可直接由决策表产生,得到了差别矩阵的几个定理.利用这些结果,给出求决策表的正区域、核、最小约简和所有属性约简的方法.理论分析和实例说明,新算法与文献[4-6]中属性约简方法所得到的约简结果相同,但效率更高.

Research on Discernibility Matrix Based on Positive Region Reduction of the Decision Table

Using the discernibility matrix,it is easy to calculate all attribute reduction of decision table.The characteristic of some created discernibility matrix in the decision table is analyzed and an improved discernibility matrix is presented.The partition U/C or positive region needs not to be calculated and the discernibility matrix can be constructed directly from the decision table.Several theorems of improved discernibility matrix are gained.Based on the above results,an algorithm using the improved discernibility matrix to calculate positive region,core,minimal reduction and all reduction of the decision table is given.Theoretical analysis and an example show that the new algorithm can have the same reduction as the algorithm given in reference ,but it requires less computational effort.