基于主动集成学习的中医智能诊断模型及构建方法  

Intelligent diagnosis model of traditional Chinese medicine based on active learning in big data

作　　者：任雪[1] 郭艳[2] REN Xue;GUO Yan(Jinan Municipal Hospital of Traditional Chinese Medicine,Jinan,250012,P.R.China;Xiyuan Hospital,China Academy of Chinese Medical Science,Beijing,100091,P.R.China)

机构地区：[1]济南市中医医院,济南250012 [2]中国中医科学院西苑医院,北京100091

出　　处：《中国循证医学杂志》2019年第9期1118-1123,共6页Chinese Journal of Evidence-based Medicine

基　　金：国家自然科学基金项目(编号:87174138)

摘　　要：作为医学和人工智能的交叉学科,智能诊断技术无论在学术界还是产业界都得到广泛关注。与基于标准化西医的智能诊断有所不同,以辨证论治为原则的中医个性化诊疗理念不同于西医,使得中医智能诊断面临有效训练样本缺失和机器学习模型失真等问题。本文提出了一种基于主动集成学习的中医智能诊断模型及其构建方法,该模型既可通过主动学习机制得到具有因人而异的个性分析能力的诊断分类器,也将多个不同机器学习的模型进行集成训练,获得更为准确的中医知识学习模型。首先将不同来源的中医病历进行大规模数据提取和组织,形成统一视图下的结构化中医数据库。然后以儿科常见病肺炎喘嗽为例,验证了基于主动集成学习的中医智能诊断模型的准确率明显高于机器学习方法,是一种新型、有效的中医诊疗机器学习模型。As an interdisciplinary subject of medicine and artificial intelligence,intelligent diagnosis and treatment has received extensive attention in both academia and industry.Traditional Chinese medicine (TCM) is characterized by individual syndrome differentiation as well as personalized treatment with personality analysis,which makes the common law mining technology of big data and artificial intelligence appear distortion in TCM diagnosis and treatment study.This article put forward an intelligent diagnosis model of TCM,as well as its construction method.It could not only obtain personal diagnosis varying individually through active learning,but also integrate multiple machine learning models for training,so as to form a more accurate model of learning TCM.Firstly,we used big data extraction technique from different case sources to form a structured TCM database under a unified view.Then,taken a pediatric common disease pneumonia with dyspnea and cough as an example,the experimental analysis on large-scale data verified that the TCM intelligent diagnosis model based on active learning is more accurate than the pre-existing machine learning methods,which may provide a new effective machine learning model for studying TCM diagnosis and treatment.

关 键 词：智能中医诊断 机器学习 主动集成学习 中医 大数据 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] R241[自动化与计算机技术—控制科学与工程]