文献类型：期刊文献

中文题名：基于有向有环图的慢性胃炎中医虚证症状选择及证候模型构建

英文题名：Syndrome model establishment and feature selection of TCM deficiency symptom of chronic gastritis based on directed cyclic graph

作者：徐玮斐[1];顾巍杰[1];刘国萍[1];王忆勤[1];颜建军[2];钟涛[2];朱生梁[3];闫秀丽[3]

机构：[1]上海中医药大学基础医学院,上海201203;[2]华东理工大学.机械电子工程中心,上海200237;[3]上海中医药大学附属岳阳中西医结合医院,上海200437

年份：2016

卷号：31

期号：11

起止页码：4491

中文期刊名：中华中医药杂志

外文期刊名：China Journal of Traditional Chinese Medicine and Pharmacy

收录：CSTPCD;;北大核心:【北大核心2014】；CSCD:【CSCD_E2015_2016】；

基金：国家自然科学基金项目(No.81270050;No.30901897;No.81173199)~~

语种：中文

中文关键词：症状;特征选择;有向有环图;证候模型

外文关键词：Symptom; Feature selection; Directed cyclic graph; Syndrome model;

摘要：目的:探讨有向有环图(DCG)在慢性胃炎中医虚证症状选择及证候模型构建中的可行性。方法:选取脾胃气虚证431例和脾胃虚寒证31例,运用粗糙集(RS)与互信息(MI)方法结合进行特征选择,然后运用DCG进行模型构建。结果:挑选出脾胃气虚证相关症状(体征)19个,脾胃虚寒证相关症状(体征)17个,筛选的相关症状基本与中医理论相符。基于筛选的症状,运用DCG方法构建证候模型,得到两种证候的分类准确率分别为脾胃气虚证74.1%和脾胃虚寒证96.0%,结果均优于朴素贝叶斯网络(NB)方法的72.1%和95.2%。结论:本研究较好地筛选了不同证候的相关症状(体征),应用考虑症状之间关系的DCG方法构建了虚证证候模型,取得了较好的分类效果,值得推广研究。
Objective： To discuss the feasibility of directed cyclic graph（DCG） in syndrome model establishment and feature selection of TCM deficiency symptom of chronic gastritis（CG）. Methods： Four hundred and thirty-one cases of syndrome of deficiency of spleen qi and stomach qi along with 31 cases of syndrome of deficient cold of spleen and stomach were selected for model establishment by DCG after feature selection by combined rough set and mutual information（RS-MI） method. Results： 19 relevant variables of the symptoms（or signs） of the spleen-stomach qi deficiency and 17 relevant variables of the symptoms（or signs） of the spleen-stomach yang deficiency were selected by the RS-MI method, which accorded with the TCM theory. The classification accuracies of both syndromes determined by DCG method were 74.1% and 96.0% respectively, which was superior to the Naive Bayes（NB） network model in terms of classification ability. Conclusion： Related symptoms of different syndromes were screened accurately in this study. Deficiency syndrome model established by DCG considering the connection between different syndromes could achieve better classification, which worth further researching and promoting.