TCM_DiffPR: knowledge graph diffusion model for personalized traditional Chinese medicine recommendation
Year: 2025
Type: article
Abstract: With the rapid development of modern Chinese medicine prescription recommendation technology based on artificial intelligence technology, the existing herb recommendation model lacks attention to the personalized attributes of the patient, ignoring the important factors of the clinical diagnosis and treatment process of Traditional Chinese medicine (TCM). To address this problem, we propose a novel Graph Knowledge Diffusion Model for Traditional Chinese Medicine Personalized Recommendation (TCM_DiffPR). Firstly, the patient's personalized attribute information is used to obtain symptom representations by means of ‘prompt fine-tuning’, and the prompt is enhanced by pre-training model with prompt-oriented comparative learning. Secondly, the algorithm combines the generative model with the data augmentation paradigm based on knowledge graph diffusion, which compensates for the misinterpretation of symptom preferences in the knowledge graph (KG) to achieve powerful knowledge learning. Finally, we introduce a collaborative knowledge graph convolution mechanism, which combines collaborative signals reflecting the symptom-herbal medication interaction patterns, to guide the knowledge graph diffusion process. In this study, experimental comparative analyses are conducted on two public datasets and a clinical dataset, and the results show that the present algorithm outperforms existing TCM prescription recommendation methods and provides an effective reference basis for clinical diagnosis and decision support.
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: G5680004860, G6259686466
OpenAlex ID: W4415324966