期刊: [Institute of Electrical and Electronics Engineers] 日期:2025-08-08卷期号:: 294-298
标识
DOI:10.1109/aiotc66747.2025.11198772
摘要
Traditional Chinese medicine (TCM) has a long history and a wide range of applications in Asian countries. An essential step to curing disease in TCM is to prescribe multiple herbs and their combinations while considering the patient's symptoms and the herbs' properties to ensure that their interactions produce the best possible results, which makes it challenging to recommend prescriptions. Lots of models are built for herb recommendation tasks. However, these methods have their limitations, such as ignoring the herb's property information and only using statistical methods to generate prescriptions. In this paper, a novel TCM prescription recommendation model based on graph convolutional networks (GCN) and neural attention mechanism is presented. Based on the advantage of the herb knowledge graph, and the fusion of multiple GCNs with a neural attention mechanism, more informative representations of herbs and symptoms are captured. The proposed model is evaluated with a TCM dataset, and the results are compared with other models. Meanwhile, ablation studies are conducted to validate the effectiveness of different components in the model. Experimental results demonstrate that the proposed model outperforms all the baselines. Concretely, compare with SMGCN, our model was 3.5%, 7.7% and 4.3% higher in Precision@5, Recall@5 and NDCG@5 respectively.