Logistic回归与决策树分类效能的ROC曲线比较  

The ROC Curves Comparing of Classification Performance between Logistic Regression and Decision Tree

作　　者：王冬燕[1] 

机构地区：[1]南京中医药大学心理学院,南京210023

出　　处：《智能计算机与应用》2014年第5期34-36,共3页Intelligent Computer and Applications

基　　金：南京师范大学哲学社会科学跨学科重大项目2010年度招标课题(KXK1010)

摘　　要：文章研究了分别利用统计学方法和数据挖掘方法,对数据进行分类和预测,具体考察了Logistic回归和决策树的分类效能,并通过ROC曲线进行检验。结果表明:两种分类方法各有利弊,没有明显的孰优孰劣,对预测变量可以进行有效的交叉验证。This paper studies the use of statistical methods, and data mining methods for data classification and prediction, specifically examines the classification performance Logistic regression and decision trees. Then the corresponding perform- ance are examined by ROC curve. The results show that： two classification methods have advantages and disadvantages, and there is no obvious one is better. These two mothods can take effective cross -validation for predicting variables.

关 键 词：LOGISTIC回归 分类回归树 ROC曲线 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]