Utilizing semantically enhanced self-supervised graph convolution and multi-head attention fusion for herb recommendation.

Journal: Artificial intelligence in medicine
PMID:

Abstract

Traditional Chinese herbal medicine has long been recognized as an effective natural therapy. Recently, the development of recommendation systems for herbs has garnered widespread academic attention, as these systems significantly impact the application of traditional Chinese medicine. However, existing herb recommendation systems are limited by data sparsity, insufficient correlation between prescriptions, and inadequate representation of symptoms and herb characteristics. To address these issues, this paper introduces an approach to herb recommendation based on semantically enhanced self-supervised graph convolution and multi-head attention fusion (BSGAM). This method involves efficient embedding of entities following fine-tuning of BERT; leveraging the attributes of herbs to optimize feature representation through a residual graph convolution network and self-supervised learning; and ultimately employing a multi-head attention mechanism for feature integration and recommendation. Experiments conducted on a publicly available traditional Chinese medicine prescription dataset demonstrate that our method achieves improvements of 6.80%, 7.46%, and 6.60% in F1-Score@5, F1-Score@10, and F1-Score@20, respectively, compared to baseline methods. These results confirm the effectiveness of our approach in enhancing the accuracy of herb recommendations.

Authors

  • Xianlun Tang
  • Yuze Tang
    Chongqing Key Laboratory of Complex Systems and Autonomous Control, Chongqing University of Posts and Telecommunications, Chongqing 400065, China. Electronic address: S220301050@stu.cqupt.edu.cn.
  • Xinran Liu
  • Haochuan Zhang
  • Xiaoyuan Dang
    School of Intelligent Engineering, Chongqing College of Mobile Communication, Chongqing 401520, China. Electronic address: dxy831110@126.com.
  • Ying Wang
    Key Laboratory of Macromolecular Science of Shaanxi Province, School of Chemistry & Chemical Engineering, Shaanxi Normal University, Xi'an, Shaanxi 710062, China.
  • Zihui Xu
    Xinqiao Hospital, Army Medical University, Chongqing 400037, China. Electronic address: zihuixu@yeah.net.