基于图论的协调多尺度决策表的最优尺度约简

Optimal scale reduction based on graph theory in consistent multi-scale decision tables

多尺度决策表是基于现实世界数据具有多尺度背景提出的一种模型。如何对多尺度决策表进行最优尺度约简是一个难题。通过构造多尺度辨识矩阵,探究辨识矩阵性质,并给出辨识矩阵与最优尺度约简的相关关系。将辨识矩阵与图论结合起来给出最优尺度约简的快速算法,最后通过数值实验验证所提出算法的有效性。

Multi-scale decision table is a model based on real-world data with multi-scale background. It is a difficult problem that how to carry out optimal scale reduction of multi-scale decision tables. By constructing the identification matrix on multi-scale, the properties of the identification matrix are put forward, and the relationship between the matrix and the optimal scale reduction is given. A fast algorithm for optimal scale reduction is presented by combining the identification matrix with graph theory. Finally, the effectiveness of the proposed algorithm is verified via numerical experiments.