CHAID-RF:基于CHAID决策树的集成学习方法  

CHAID-RF:Ensemble Learning Method Based on CHAID Decision Tree

作　　者：聂斌[1] 靳海科 李欢 陈裕凤 张玉超 郑学鹏 NIE Bin;JIN Haike;LI Huan;CHEN Yufeng;ZHANG Yuchao;ZHENG Xuepeng(College of Computer Science,Jiangxi University of Chinese Medicine,Nanchang 330004,China)

机构地区：[1]江西中医药大学计算机学院,江西南昌330004

出　　处：《现代信息科技》2024年第17期28-35,42,共9页Modern Information Technology

基　　金：国家自然科学基金项目(82260849,61562045);江西省教育厅科技计划研究项目(GJJ211256);江西中医药大学校级科技创新团队发展计划(CXTD22015)。

摘　　要：针对卡方自动交互诊断(CHAID)决策树易过拟合的问题,提出CHAID随机森林方法(CHAID Random Forest,CHAID-RF)。该方法采用随机采样、随机选择特征以及集成的策略,将CHAID决策树作为基分类器,形成CHAID-RF。为了验证CHAID-RF的有效性,选取CART、CHAID、SVM、RF作为对比算法,以准确率、加权查准率、加权查全率、加权F值作为分类模型评价指标,以均方根误差作为回归模型评价指标,采用10个分类数据集和7个回归数据集进行验证。实验结果表明CHAID-RF可行有效。Aiming at the problem that CHAID Decision Tree is easy to overfitting,CHAID-RF is proposed.In this method,CHAID Decision Tree is used as the base classification to form CHAID-RF by random sampling,random feature selection and integration strategies.CART,CHAID,SVM,and RF are selected as the comparison algorithm to verify the effectiveness of CHAID-RF,accuracy,Weighted Precision Ratio,Weighted Recall Ratio,and Weighted F-measure are used as evaluation index of classification model,and Root Mean Square Error is used as evaluation index of regression model,10 classification data sets and 7 regression data sets are used for validation.The experimental results show that CHAID-RF is feasible and effective.

关 键 词：CHAID 随机森林 CHAID-RF 分类 回归 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]