基于ROUGH集的决策树测试属性选择方法

测试属性的选择直接影响到决策树中结点的个数甚至是深度,因此如何选择测试属性是研究的一个热点。本文主要介绍了粗集理论的方法。通过比较我们会发现,在单变量决策树的构造上,粗集理论中属性重要性的方法计算量较小,而多变量决策树充分考虑了条件属性间的相关性,因此通过求解信息系统的相对核从而减少决策树结点的个数。

How to Choose Test Attributes in a Decision Tree Based on Rough Set

How to choose test attributes, plays a role in constructing the decision tree. Therefore it has become an attractive research direction. Now there are many different methods to solve it. In this paper the two basic methods are introduced such as information entropy and rough sets. Comparing those methods, the method of the important attributes in rough sets is understood easily and decreases the computation in the way of constructing the single vary decision tree. But the relative of the condition attributes is ignored. On the contrary, the core(s) is found at first using rough sets method by considering the relation of condition attributes. The method of the many vary decision tree puts the core(s) into the root node, so it reduces the nodes in a tree.