连续属性值的整体离散化

Global discretization of continuously valued attributes

给出一种基于超立方体聚类的连续性值整体离散化方法 .该方法先对训练例子进行超立方聚类 ,然后通过聚类区域在每个连续属性轴上的必要投影区间推导出每个连续属性的离散化划分点 .实验表明本方法不仅能显著减少离散化划分点和归纳规则数 ,而且能提高分类精度 .

The global discretization of continuously valued attribute is one of the essential techniques in applying symbolic inductive learning algorithms. Based on hyper cube clustering, a global discretization approach has been provided. Experimental results indicate that global discretization approach can significantly decrease the number of discretization cut points and the number of rules, but increase the accuracy of the classifiers.