Scholarlink

<p>Artificial Intelligence (AI) is increasingly favored for traditional Chinese medicine (TCM) prescription recommendation tasks. However, existing methods often overlook personalized patient attributes and fail to provide reasonable explanations for the recommended herbs, making them inconsistent with clinical TCM processes and unsustainable for recommender systems. To address these challenges, we propose an interpretable method for personalized TCM prescription recommendations aimed at ensuring health and well-being (TCM-IPRW). Our approach utilizes graph neural networks (GNN) and attention mechanisms, coupled with a manually constructed TCM knowledge graph, and incorporates additional proximity z-score metrics to enhance model interpretability. Furthermore, we concentrate on ensuring the technical and social sustainability of our recommendation approach. Extensive experiments conducted on two publicly available datasets and one clinical dataset validate the effectiveness of our method, offering fresh insights into TCM clinical practice while advancing the modernization and sustainability of TCM diagnosis and treatment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Interpretable methods for personalized recommendation of traditional chinese medicine

  • 期刊论文

  • 发表日期：

    2025年9月30日

查看全文
  • Chaobo Zhang,
  • Long Tan

摘要

Artificial Intelligence (AI) is increasingly favored for traditional Chinese medicine (TCM) prescription recommendation tasks. However, existing methods often overlook personalized patient attributes and fail to provide reasonable explanations for the recommended herbs, making them inconsistent with clinical TCM processes and unsustainable for recommender systems. To address these challenges, we propose an interpretable method for personalized TCM prescription recommendations aimed at ensuring health and well-being (TCM-IPRW). Our approach utilizes graph neural networks (GNN) and attention mechanisms, coupled with a manually constructed TCM knowledge graph, and incorporates additional proximity z-score metrics to enhance model interpretability. Furthermore, we concentrate on ensuring the technical and social sustainability of our recommendation approach. Extensive experiments conducted on two publicly available datasets and one clinical dataset validate the effectiveness of our method, offering fresh insights into TCM clinical practice while advancing the modernization and sustainability of TCM diagnosis and treatment.