基于R语言rpart包与party包的分类树建模研究  

作　　者：徐万佶 曾庆[1] 

机构地区：[1]重庆医科大学公共卫生与管理学院,重庆400016

出　　处：《科技资讯》2018年第11期20-22,共3页Science & Technology Information

摘　　要：目的比较R语言中rpart包与party包所构建决策树的不同。两个不同的包所构建的模型可推广应用于其他领域的决策树分类问题。实验过程中涉及的数据预处理、分类树建模、模型可视化、测试的思路与方法,也可借鉴应用于其他领域的分类树建模预测工作。方法对R语言内置数据集iris应用分类回归树(classification and regression tree,CART)方法与条件推断决策树,并运用R语言实现并建立决策树模型。结果 rpart包所得决策树模型与party包所得决策树模型在预测iris测试集的准确率均为96.7%。结论 rpart包所得到的决策树与party包所得到的决策树在对iris测试集的预测效果上无差异。Objective To compare the dif ferences between the rpart and party packages in the R language.The model constructed by two different packages can be applied to decision tree classif ication in other f ields.The data preprocessing,classif ication t ree modeling,model visualization,and t esting ideas and methods involved in t he experiment process can also be used for modeling and prediction of classif ication trees applied to other f ields.Methods The classif ication and regression tree(CART)method and conditional inference tree were applied to the built-in data set iris of the R language,and a decision tree model was implemented using R language.Results The accuracy rate of the iris test set was 96.7%for the decision tree model obtained from the rpart package and the decision tree model obtained from the party package.Conclusion There is no difference in the prediction effect of the iris test set between the decision tree obtained by the rpart package and the decision tree obtained by the party package.

关 键 词：分类决策 树R语言 rpart包 party包 

分 类 号：C819[社会学—统计学]