面向在线医疗平台的医生推荐方法  

Doctor recommendation approach for online healthcare platforms

作　　者：刘珂 刘盾[1] 孙扬 沈蓉萍 LIU Ke;LIU Dun;SUN Yang;SHEN Rongping(School of Economics and Management,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学经济管理学院,四川成都610031

出　　处：《智能系统学报》2025年第1期206-218,共13页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(62276217,61876157);四川省青年科学基金项目(2022JDJQ0034);中央高校基本科研科技创新项目(2682024KJ005,2682024ZTPY021).

摘　　要：近年来,随着智慧医疗的日益普及,在线医疗平台已逐步发展为满足大众基本医疗需求的重要渠道。为患者推荐合适的医生是在线问诊中的一个重要过程,优化推荐能力不仅可以提高患者的满意度,还能够推动在线医疗平台的发展。与传统推荐系统不同,医生推荐领域受到隐私保护限制,无法查看患者曾经的诊疗历史,因此模型训练时仅能利用每位患者最近一次的就诊记录,面临严峻的数据稀疏问题。同样,模型预测时也仅能根据患者当前的疾病描述文本进行推荐,而由于患者对疾病描述方式的差异性,模型对不同患者的推荐能力也存在差异,这会使部分患者的需求无法得到满足,进而影响模型整体的推荐能力。基于此,本文提出了一种基于数据增强的医生推荐方法(sequential three-way decision with data augmentation,STWD-NA),通过引入不匹配的医患交互信息扩充训练数据,并利用序贯三支决策的思想训练模型。具体来说,该方法由两部分组成:一方面引入了不匹配交互信息的方法,以缓解训练冷启动问题;另一方面,提出了一种基于序贯三支决策的训练算法,以动态调整模型训练时的关注度。最后,通过好大夫平台上的真实数据集验证了本文所提STWD-NA方法的有效性。Online healthcare platforms have become increasingly important in meeting the basic medical needs of the public,especially with the growing popularity of smart healthcare.A crucial step in the online consultation process is helping patients find a doctor with relevant expertise,as this not only enhances patient satisfaction but also fosters the development of online healthcare platforms.Unlike traditional recommendation systems,doctor recommendations are subject to privacy protection,and historical records for each patient cannot be accessed.As a result,models can only utilize the most recent consultation records for training,leading to severe data sparsity issues.Similarly,during prediction,recommendations are made solely based on the patient’s current disease description.However,different patients describe their conditions in different ways,which results in varying recommendation effectiveness across patients.This may fail to meet the needs of some,thereby affecting the overall performance of the recommendation system.Along this line,in this paper,we proposed a novel method called Sequential three-way decision with data augmentation(STWDNA),which combined both matching and mismatched interaction information for doctor recommendation to expand training data.Specifically,this novel method consisted of two parts.On the one hand,it proposed a method to integrate the mismatched interaction information to alleviate the cold-start problem during training.On the other hand,an algorithm was proposed based on the idea of sequential three-way decisions to dynamically adjust the model’s attention during the training process.Evaluation based on real-world dataset haodf.com demonstrates the utility and the effectiveness of the proposed method.

关 键 词：在线问诊平台 医生推荐 序贯三支决策 多粒度 数据增强 负样本 负采样 数据稀疏 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]