连续量决策信息表的离散化问题研究

Discretization question research on continuous decision information tables

在粗糙集理论中,知识是以表格的形式表达的。当用粗糙集算法从连续量决策信息表中提取规则时,首先要对其进行离散化处理。针对连续量决策信息表离散化过程中存在的问题,本文中提出了2个定理并进行了证明。定理表明:在考虑决策信息表中属性值之间不可分关系的条件下,若决策信息表的值发生变化,则离散化结果必然发生变化。所以由单一样本构成的连续量决策信息表所得到的离散化结果不能用于实际的连续量系统中。

The knowledge is represented into the table form in rough set theory. Discretlzatlon ot continuous decision information tables （CDIT） is necessary before extracting rules from it applying rough set algorithm. Aiming at existent question in the process of discretization of the CDIT, two theorems are proposed and proved in this article. The theorem state that： in considering the indiscernibility relation between attribute values under the conditions of the DIT, if the values of the DIT changes, then the discretization result will inevitably change. So the discretization result obtained from a CDIT composed by single data sample can not be used in practical continuous variable system.