Traditional Chinese Medicine Prescription Recommendation Technology Using Individual Characteristics

Traditional Chinese Medicine（TCM）making needs complex knowledge and experience.TCM has a variety of efficacy and usage.There is complex knowledge between TCM prescription and herbs.The influence of herb alone and the combination of two herbs can produce positive and negative effects.TCM diagnosis and treatment not only need to consider symptoms,but also other individual characteristics of patients.Patients with different individual attributes may have different symptoms when they suffer from the same diseaseHowever.The current TCM recommendation model has not yet considered the complete individual characteristics of patients.In addition to the compatibility of traditional Chinese medicine,the dosage selection is also one of the difficulties in the prescription making.In the field of traditional Chinese medicine,there is a gap in the analysis of the dosage of single-prescription traditional Chinese medicine for various syndromes.The lack of individual attributes leads to the lack of personalization of recommendation model,and the performance is relatively insufficient.Meanwhile,the dose prediction under specific prescriptions for a variety of disease problems needs to be explored urgently.Firstly,aiming at the lack of individual attributes in the prescription recommendation task,this paper studies the traditional Chinese Medicine recommendation model integrating the individual characteristics of patients.Further,we have preliminarily explored the dosage range and accurate dosage prediction in the prescription.Based on the above work,a patient oriented visual analysis system of traditional Chinese medicine is proposed.The main work is as follows:·Herb recommendation model incorporating individual characteristics Aiming at the lack of individual attributes in the prescription recommendation,we express the patient’s individual characteristics based on the heterogeneous graph,that is,expresses the relationship between the patient’s individual attribute,symptom and diagnostic,so as to extract interactions in prescription during diagnosis inducing.In view of the defect that the existing model does not distinguish between multiple entity types in graph learning,the correlation graph of individual characteristics of patients and herbs is constructed and the type-aware recommendation model is designed to distinguish the influence of different medical entities on the selection of traditional Chinese medicines,and to learn the relationship between traditional Chinese medicine and diagnosis.This is helpful for downstream traditional Chinese medicine recommendation task.Specifically,this paper builds a node type-aware Chinese medicine recommendation model based on multi-GCN to learn representations for patients and herbs and emphasizes individual characteristics to solve the problems of lacking individual characteristics and not distinguishing different entities.Through experiments,it is proved that modeling individual patient characteristics can improve the quality of patient feature representation.Also,distinguishing different entity node types in graph convolutional neural networks improves the effectiveness of message passing.·Dosage range and accurate dosage prediction To solve the problem of dose prediction in prescription recommendation,this paper has launched the first exploration of the task of predicting the dosage of traditional Chinese medicine for various diseases.This paper proposes a dosage range prediction task and a dosage prediction task for different diseases,with individual patient characteristics and prescription composition characteristics.Specifically,the prediction of the dosage range is based on the classification model.The accurate dosage prediction is based on the multiregression model integration framework are respectively proposed.Experiments and analysis prove the effectiveness of our design to use the individual characteristics and prescription information.·TCM Prescription analysis system For TCM patients,we build a prescription recommendation and analysis system.Combined with historical information,we statistically analyze the prescriptions held by patients.The individual characteristics of patients are integrated,and corresponding prescriptions and dosage references are provided based on our herb recommendation and dose prediction model.It vividly displays the relevance of drugs in a single prescription in a visual form to provide convenience for users.And in the form of the knowledge graph,the relationship with other medical entities are visually expanded for patients and other users.