决策树算法是一种采用分治策略的自顶向下的归纳算法,传统的决策树算法往往是基于信息论度量的.文章以粗糙集合理论中的区分观点为基础,提出了两种新型的属性选择判据：区分度和区分价值.实验结果表明,采用区分价值的属性选择策略所生成的决策树要明显优于基于熵的属性选择策略. The decision tree algorithm is a top-down induction algorithm that adopts divide-and-conquer strategy, and the traditional decision tree algorithm is often based on information theory.On the basis of the distinction viewpoint in rough set theory, this paper proposes two new types of attributes Choice Criteria: Distinguishing Degree and Distinguishing Value.Experimental results show that the decision trees generated by using attribute-based selection strategies are obviously superior to attribute-based selection strategies based on entropy.