文献类型：期刊文献

中文题名：基于极值随机森林的慢性胃炎中医证候分类

英文题名：Syndrome Classification of Chronic Gastritis Based on Extremely Randomized Forest Algorithm

作者：颜建军[1];胡宗杰[1];刘国萍[2];王忆勤[2];付晶晶[2];郭睿[2,3];钱鹏[2]

机构：[1]华东理工大学机械与动力工程学院,上海200237;[2]上海中医药大学四诊信息综合实验室,上海201203;[3]上海中医药大学交叉科学研究院,上海201203

年份：2017

卷号：43

期号：5

起止页码：698

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：北大核心:【北大核心2014】；CSCD:【CSCD_E2017_2018】；

基金：国家自然科学基金(81270050;81302913;30901897;81173199)

语种：中文

中文关键词：证候分类;极值随机森林;可解释性;慢性胃炎;决策树

外文关键词：syndrome classification; extremely randomized forest; interpretability ; chronic gastritis; decision tree

摘要：大多数机器学习算法能得到较好的分类效果,但模型却无法解释;而随机森林等模型有良好的可解释性,却无法处理中医数据中兼证的情况。本文利用极值随机森林算法对慢性胃炎中医数据进行证候分类研究,其中决策树的叶节点能输出多个标签,通过加权机制综合分量来处理兼证问题。与已有多标记学习算法和C4.5、CART等基于决策树的算法进行比较,实验结果表明,极值随机森林算法无论在6个证型的分类准确率上,还是在多标记评价指标上都具有更好的效果,而且模型中得到的规则基本符合中医理论。
Syndrome differentiation and treatment,which is the essence of traditional Chinese medicine(TCM),contain abundant rules.The majority of machine learning algorithms can obtain good classification accuracy,but these models are difficult to be explained.The models established by random forests have great interpretability,while these models cannot deal with multi-syndrome that patients may simultaneously have more than one syndrome in TCM.In this paper,syndrome classification for Chronic Gastritis(CG)is researched by using extremely randomized forest(ERF)algorithm,and compared with state-of-the-art multi-label algorithms and the tree-based algorithms(such as C4.5,CART).The experimental results show that ERF algorithm has better performance than other algorithms in the classification accuracy of every label and the six evaluation metrics of multi-label learning.The rules obtained in the model are basically in accord with TCM theory.