叶枝分辨的决策树研究  

Research on decision tree based on the distinguishment of leaves and branches

作　　者：刘军[1] 

机构地区：[1]南京工业大学电子与信息工程学院,江苏南京210009

出　　处：《计算机工程与设计》2013年第9期3245-3249,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(60673185);教育部留学回国人员科研启动基金项目(200711108)

摘　　要：为了解决当前构建决策树算法中准确度与计算量之间的矛盾,根据粒计算理论,提出了利用叶枝粒之比的决策树构建算法:将条件属性按属性值划分为若干基本粒作为基本分配单位(叶重量),而将基本粒所对应的决策属性值的类别数量作为基本接受单位(枝数量),直接利用叶枝粒比判断确定划分属性,由所选划分属性自顶向下构造决策树。通过理论与实例分析结果表明了该算法简洁且准确性高。To solve the contradictions between accuracy and computation in the current construction of decision tree algorithm, a decision tree construction algorithm is proposed based on the grain ratio of the leaves and branches according to the grain compu- ting theory. The condition attribute by the values are divided into a number of basic grains as the allocation unit （leaf weight）, and the number of these grains corresponding to the classes of the decision attributes are as the basic accepted unit （the quantity of branches）, the partition attribute is directly determined by the grain ratio of the leaves and branches and the decision tree is built with the selected attribute from top to down. Through theoretical and experiment analysis, this algorithm is concise and ac- curate.

关 键 词：粗糙集 决策树 粒分辨关系 叶枝粒比 划分属性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]