摘要
运用 Rough集理论处理决策表时 ,要求决策表中的值用离散数据表达。文中形式化描述了离散化问题 ,研究了决策表的属性重要性 ,提出计算条件属性重要性的新方法 ,然后给出基于属性重要性的决策表离散化算法。该算法按照属性重要性从小到大的顺序对每个属性进行离散化 :从可辨别矩阵中得到冲突样本 ,将冲突样本的属性值作为断点的上下界 ,用所有这样的断点化简后得到的断点集对该属性进行离散化处理。
In this paper, the discretization in decision tables is first summarized. By analyzing the importance of attributes in decision tables, a new method is put forward to compute it. Based on the attributes' importance, an algorithm is proposed to discretize multi valued attributes in decision tables. According to the discernibility matrix, the value region of a break point can be obtained, and discretization can be started with these break points.
出处
《系统工程理论方法应用》
2003年第2期141-145,共5页
Systems Engineering Theory·Methodology·Applications
