一种连续属性离散化的新方法

A New Method for Discretization of Continuous Attributes

传统的粗糙集理论只能处理离散属性,所以在对决策表进行处理之前,必须对决策表中的连续属性进行离散化。提出一种新的相容离散化算法,从决策属性值出发,对决策属性值相同的实例在空间上进行合并,并用例子说明该算法的离散化过程。最后分析了该算法的复杂度。该方法能有效地离散连续属性决策系统,结果表明,采用本算法进行离散化能得到合理的断点结果。

Because traditional Rough Sets theory can only deal with discrete attributes, continuous attributes must be converted into discrete attributes before we cope with decision table. In this paper, a new algorithm of consistent discretization in decision table based on rough sets theory is presented for the continuous attributes. The discretization algorithm is performed by selecting the same decision attribute value from the discretized attributes,and classfy the unions as a class. This algorithm is effectively tested by an example. The method’s complexity is analysed and compared with some other discretization algorithms. Results show that using this approach can get reasonable intervals.