一种基于改进DRNN网络的决策树构建方法  

A method of constructing decision tree based on improved DRNN network

作　　者：郭娜[1] 田亚菲[1] 郝洁[1] 贾存丽[1] 

机构地区：[1]兰州大学信息科学与工程学院,甘肃兰州730000

出　　处：《软件》2010年第11期8-11,共4页Software

摘　　要：决策树是数据挖掘和归纳学习的重要方法。本文介绍了ID3算法,ID3算法存在着倾向于取值较多属性的缺点;神经网络也可以用来分类,但是神经网络不易于分类规则的提取。在遗传算法优化的DRNN网络的基础上,提出了使用差量法构建决策树的方法。该方法既具有神经网络分类的高精度,而且分类规则比较直观。实验数据分析表明,本文提出的方法更加接近实际情况,能够更好的进行预测和分类。Decision tree is an important method for data mining as well as induction learning. ID3 algorithm is introduced in this paper. The disadvantage of ID3 algorithm has an tendency to choose attributions which has more values; neural network, which is not easy to extract the claasification rules, can also be used to classify things. Based on DRNN network which is optimized by Genetic algorithm, the use of the different method to build decision tree is proposed. It has the high-precision that neural network classifier does, meanwhile, its classification rules are intuitive. Experimental data shows that the proposed method is more practical and able to predict and classify things in a better way.

关 键 词：数据挖掘 决策树 ID3算法 神经网络 差量法 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]