Machine Learning Based Recommendation System For Traditional Chinese Medicine (TCM) Formulas

As a part of the five thousand year Chinese civilization,traditional Chinese medicine(TCM)is of great value.However,as an empirical discipline,its inheritance gradually begins to weaken after entering the era of electronic information.It can only rely on teaching and apprenticeship for transmission.The omission,error and a small part of uncertainty in the transmission process are inevitable for the inheritance of TCM knowledge No wastage.Therefore,it is the best choice to integrate TCM experience and TCM knowledge into a carrier that will not be affected by non objective factors.The traditional Chinese medicine prescription writing,need clear judgment about the patient’s existing disease,syndrome in order to make a correct summary,and get the prescription of patients.This process needs a long time to judge and adjust.Secondly,the existing software does not provide a good auxiliary method.Based on the above problems,this paper mainly studies how to adopt the machine learning approach to make recommendations on TCM Formulas,through entity extraction techniques,association rule mining methods,to form a reliable data base and borrow from the then and then more popular programming languages and frameworks,and to construct a system to make reliable recommendations on the syndromes and formulas of disorders.In view of the above problems,the main research contents of this paper are as follows:1)As there is no standard open data set for the correspondence between diseases and prescriptions in TCM,this paper constructs a total of 1982 data sets through prescriptions and corresponding diseases provided in Synopsis of the golden chamber.This paper proposes a named entity extraction method,which processes the basic data into the form of mathematical vector through word2 vec model,and then inputs it into a comprehensive model of superposition of bi-directional long short term memory(BI LSTM)neural network with coupled forgetting gate and memory gate and conditional random field(CRF).2)Put forward an improved mining model of association rules with multiple machine learning methods,which is proposed to mine association rules for terminology entities existing in Synopses of the Golden Chamber,and to analyze the extracted association rules among entities.In addition,compared with the traditional Apriori algorithm,the association model based on the relationship threshold mechanism derived from the context relationship calculation has a better performance in the mining of association rules when the amount of data is increasing.The mining entity strong association rules are stored in the My SQL database,so as to provide the main recommendation items for the card recommendation system.3)Borrowing the basic data generated from the previous study,the association rule as the underlying support,the Classification Based Association rule(CBA)algorithm joint react framework,Java Script and other programming technologies were adopted to realize an interactive friendly and comprehensible machine learning based recommendation system for TCM Formulas.