分类问题是数据挖掘和机器学习中的一个核心问题。现在的数据都是多维的，包含非常多的属性，而其中某些属性甚至是冗余的，会干扰人们正确而简洁地决策，为了得到最大程度的分类准确率，决策树分类过程中，非常关键的是决策属性的选择。常见的决策属性选择方法可以分为相关分析方法、主成分分析方法等。分析了粗糙理论的属性约简方法，提出了基于粗糙理论的属性约简在决策树中的应用，并产生多个相对约简的选择问题。

　Classification is a core problem in data mining and machinery learning. Nowadays data are usually multi-dimensional, which contain so many properties. However, some properties may be redundant and interfere with the correct and clear decisions. In order to obtain the maximum classification precision, it is so important to select the proper decision tree in the process of classification of decision tree. The common methods of decision property selection include relevant analytical method, componential analysis methods and so on. This paper analyzed the method of property reduction about rough theory and applied it to decision tree, which can produce multiple choice problems of relative reduction.