面向属性值分类的决策树研究  

Research on Decision Tree for Attribute Classified by the Value

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：《成都信息工程学院学报》2013年第4期347-352,共6页Journal of Chengdu University of Information Technology

基　　金：国家自然科学基金资助项目(60673185)

摘　　要：为了提高当前决策树构建的准确度,降低计算复杂度,以粒分辨关系为理论基础将属性按其值划分为若干子集(粒)作为基本判决单位,提出叶、枝划分集的逻辑式判定算法,将条件属性粒的长度作为可分配的树叶量,而其对应的决策粒作为可接受的树枝量,用叶、枝比例的强度判定该属性粒的分辨力,进而从全局分辨关系直接判定出划分属性,由顶向下进行决策树的构建。实例分析的结果表明,算法简洁且精准。In order to improve the accuracy and reduce the computation complexity in the construction of the decision tree, according to theoretical principle of granulation distinguish, the attribute is divided by the value into several subsets（grain） as the basic judgment unit, and the logical expression decision algorithm is put forward based on partition sets of leaves and branchses. The condition attribute grain length can be assigned as amount of the leaves, and the grain of the corresponding decision-making is the acceptable branch quantity, the resolution of attribute grain is determined by ratio strength of leaf and branch, then the division attribute is directly determined by the global resolution relationship, the decision tree is constructed from top to down. The example analysis results show that the algorithm is simple and accurate.

关 键 词：计算机应用技术 数据挖掘 粗糙集 决策树 粒分辨关系 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]