基于决策分类熵的决策树构造算法及应用  

Application of decision tree construction algorithm based on decision classify-entropy

作　　者：董广[1] 王兴起[1] 

机构地区：[1]杭州电子科技大学计算机学院,杭州310018

出　　处：《计算机应用》2009年第11期3103-3106,共4页journal of Computer Applications

摘　　要：为了更好地完成金融数据集上的分类挖掘任务,以粗糙集理论为基础提出决策分类熵的概念,进而以属性的决策分类熵为属性分裂度量提出基于决策分类熵的决策树构造算法,并针对过拟合问题提出一种抑制参数来实现树规模的良好控制。实例分析及金融数据集上的实验表明:相比经典的C4.5决策树算法,新算法能够较好地克服其缺点和不足,构建更优的决策树,能够更好地完成分类任务。In order to better complete the task of classification mining on financial datasets, decision classify-entropy concept was put forward based on the rough set theory; and based on this concept, a novel decision tree construction algorithm was proposed. To overcome over-fitting, inhibiting factor was introduced to control decision tree construction. The case study and experimental results in financial datasets show that, compared with the classical C4. 5 algorithm, the new algorithm can resolve the drawbacks of the traditional algorithm and could construct a suboptimal decision tree effectively. The application in financial field also proves that the new algorithm can finish the objective task much better.

关 键 词：决策树 粗糙集 决策分类熵 抑制参数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]