基于机器学习的互联网诊疗医患智能匹配研究  

Research on Two-stage Doctors-patients Intelligent Matching on Internet Medical Platforms Based on Machine Learning

作　　者：张利江[1] 桂亮 赵晓晓 ZHANG Lijiang;GUI Liang;ZHAO Xiaoxiao(School of Management,Xinxiang Medical University,Xinxiang 45300,Henan,China)

机构地区：[1]新乡医学院管理学院,河南省新乡市453000

出　　处：《中国卫生信息管理杂志》2023年第1期109-115,共7页Chinese Journal of Health Informatics and Management

基　　金：河南省哲学社会科学一般项目《互联网医疗医患信息协同结构解析与信息智能匹配研究》(项目编号:2020BZH004);河南省科技攻关项目《互联网医院患者画像构建及医生智能推荐策略研究》(项目编号:212102310499)。

摘　　要：目的互联网医疗服务平台汇集了大量医生,患者难以从中找到与自己就医需求高度匹配的医生,探索医患智能匹配的方法,实现患者精准择医。方法将患者咨询记录标题作为文本挖掘对象,采用机器学习方法,设计了针对“医生聚类”和“医患匹配”的机器学习模型,基于LDA主题模型进行医生聚类,再利用咨询记录训练出代表各聚类中心类医生的朴素贝叶斯分类器,判断患者咨询问题是否与此类医生的医学背景匹配。结果数值实验显示,该医患智能匹配方案具有较好的性能,测试一组实验数据中匹配准确度最高为99.3%,测试二组实验数据中匹配准确度最高为97.8%。结论基于患者咨询记录标题,综合LDA和朴素贝叶斯分类技术可提高患者互联网就医过程中的择医效率。Objective In online medical treatment,it is difficult for patients to find doctors who are highly matched with their own medical needs from massive information.This paper aims to explore the method of intelligent matching of doctor-patient information on internet medical platform and realize patients’precise choice of doctors.Methods The title of patient consultation record(TPCR)is taken as the object of text mining.Latent Dirichlet Allocation(LDA)topic model and a Naive Bayesian Classifier are used to judge whether the patient consultation questions match the medical background of such doctors.Results Our numerical experiments show that the doctor-patient intelligent matching scheme has good performance.For the first test group collection,the highest accuracy is 99.3%,and for the second one that is 97.8%.Conclusion Based on the title of patient consultation record,integrated of LDA and naive Bayes classifier,our method can improve the efficiency of medical selection in the process of patients seeking medical treatment on the internet.

关 键 词：互联网医疗 LDA主题模型 朴素贝叶斯分类器 机器学习 患者咨询记录标题 

分 类 号：R-39[医药卫生] R319