文献类型：期刊文献

中文题名：基于知识图谱的中药推荐方法

英文题名：Traditional Chinese medicine recommendation method based on knowledge graph

作者：李春雨[1];李燕[1];刘悦悦[1];刘雪丽[1]

第一作者：李春雨

机构：[1]甘肃中医药大学信息工程学院,甘肃兰州730100

第一机构：甘肃中医药大学信息工程学院（教育技术中心）

年份：2025

卷号：34

期号：6

起止页码：731

中文期刊名：云南民族大学学报(自然科学版)

外文期刊名：Journal of Yunnan Minzu University(Natural Sciences Edition)

基金：中国高校产学研创新基金(2021LDA09002);甘肃省研究生创新创业基金(2023CXZX-770).

语种：中文

中文关键词：中药推荐;辅助诊疗;图卷积网络;知识图谱

外文关键词：traditional Chinese medicine recommendation;auxiliary diagnosis and treatment;graph convolutional network;knowledge graph

摘要：为解决中药推荐问题,在推荐中引入更多的中医药知识和方法,提出基于知识图谱的中药推荐模型.首先构建症状-中药知识图谱,将其应用于预训练阶段,更好地保留了图结构信息,其次,根据在处方中的共现频次构建症状-症状图和中药-中药图,挖掘同质节点之间隐含的规律,获得更全面的节点嵌入.在模型的预测层,将多个症状的集合看作一个整体,利用多层感知机提取症状集的特征,更好地表示多个症状之间的联系.实验结果表明,该模型在3种评价指标上均取得较好效果.
To solve the problem of traditional Chinese medicine recommendation,a knowledge graph-based recommendation model is proposed by incorporating more traditional Chinese medicine knowledge and methods into the process.First,a symptom-traditional Chinese medicine knowledge graph was constructed and applied in the pre-training stage to better preserve graph structure information.Then,based on the co-occurrence frequency in prescriptions,a symptom-symptom graph and a traditional Chinese medicine-traditional Chinese medicine graph were built to mine hidden rules among homogeneous nodes and obtain more comprehensive node embeddings.Furthermore,in the prediction layer of the model,a set of multiple symptoms is regarded as a whole,and a multi-layer perception is used to extract features from the symptom set,thereby better representing relationships among symptoms.Experimental results show that the model achieves favorable performance across three evaluation metrics.