决策树是归纳学习的重要形式 ,建造高质量的决策树的关键是选择合适的属性 .本文针对ID3算法对属性间的相依性强调不够等问题 ,利用粗糙集理论 ,提出一种新的启发式函数———分辨率构造决策树 .分辨率本质上是相关属性的组合 ,但不是简单的属性合取 .它不仅考虑了属性之间的依赖性 ,还兼顾了分类的种数 .大量实例表明 ,本文的方法明显优于ID3算法 . Decision tree is an important form of inductive learning, and the key to build a high-quality decision tree is to select the appropriate attributes.In this paper, ID3 algorithm for the interdependencies between attributes not emphasized enough and so on, the use of rough set theory, a new heuristic Function --- Resolution Structure Decision Tree. Resolution is essentially a combination of related attributes, but not a simple attribute combination. It takes into account not only the dependencies between attributes but also the number of categories. A large number of examples show , This method is obviously better than ID3 algorithm.