Research And Application Of Auxiliary TCM Prescription Recommendation System

Traditional Chinese Medicine(TCM)science is a set of integrated and peculiar theory systems,which is formed through the accumulation,inheritance,summarization and innovation of experience in disease prevention and treatment.Doctors of TCM make prescriptions following three basic steps: Firstly,diagnose the dieases.To observe and understand the patient’s diseases through looking,listening,asking and feeling the pulse;Secondly,try to differentiate the syndrome.After the obtaining of the objective data through the ”four diagnostic methods”,the theory of TCM science was used to analyze the symptoms and make conclusions in syndrome differentiation;Finally,make TCM prescriptions.The treatment principle of TCM is a general theory that gradually accumulated by the working people in long-term medical practice.It is closely related to diagnosis and syndrome differentiation and constitutes a set of complex theory systems containing TCM theory,law,prescription,and medicine.Besides making the diagnosis of the disease and the accuracy of making prescriptions depend heavily on the doctors’ experience,the cost of diagnosis and making prescriptions is very expensive.Our main work is to develop the algorithms for the selection of TCM prescription and to implement an auxiliary TCM prescription recommendation system by using these algorithms.The system can assist medical personnel to make prescriptions according to the patients’ conditions,maximizing the labor saving.The main contents of this thesis includes:1)As the data of TCM prescriptions has the characteristic of ”semi-structured”,the preprocessing of data was studied.The method of constructing vocabulary lists standardizes the data of TCM prescriptions,which provides the necessary conditions for the subsequent work.2)To measure the similarities between the standard efficacies.Based on the correlation between the symptoms and the efficacies,the SimRank algorithm is used in this thesis to obtain the similarity between the efficacies.3)The construction of neural network through feedbacks.Firstly,a derivative LDA model is used in reduction of the standard efficacies.Secondly,based on the characteristics of the prescription data,that is,the ”equivalence” between prescription symptoms and prescription efficacies,two neural networks are constructed,with symptoms and efficacies as input and the ”differences” between the two inputs as training targets.After training the model,this system makes prescriptions by transversing the classic TCM prescriptions.In view of the slow speed,a faster neural network model is proposed for making prescriptions.4)The designation and implementation of an auxiliary TCM prescription recommendation system.The system separates the display interface from the process of making prescriptions by using MVC design pattern.After the doctor determines the specific symptoms of patients through inquiries,the system automatically makes prescriptions based on these symptoms.According to the actual tests,the system makes prescriptions automatically with convenience and with accuracy.