Research On Knowledge Graph Learning And Reasoning For TCM Prescription Of Chronic Kidney Disease

Chronic kidney disease is a common clinical disease,and its disease is hidden.The course of disease is longer,and early diagnosis and effective treatment is of great significance for effective prevention and treatment.Modern medicine for the treatment of chronic kidney disease is limited,and it is mainly through the regulation of diet,diuretic,hypotension,lipid-lowering and other related risk factors control.For a long time,Chinese medicine has its unique diagnosis and treatment in the treatment of chronic kidney disease,and has achieved remarkable results.Traditional Chinese medicine believes that the main characteristics of chronic kidney disease are deficient root and excessive superficiality,early spleen and kidney deficiency,late qi and blood yin and yang deficiency,so the treatment pays attention to both root and superficiality.Deficiency and excess regulate each other.At present,Chinese medicine practitioners with rich experience are rare,and limited to the traditional Chinese medicine inheritance mode,doctors who lack rich clinical experience can not meet the urgent needs of the public for the diagnosis and treatment of high-quality chronic kidney disease.It is urgent to introduce modern information technology to assist Chinese medicine diagnosis and treatment experience to be passed on effectively and make decisions.This paper aims at the traditional Chinese medicine case of chronic kidney disease,and constructs the knowledge graph of chronic kidney disease.It combines the modern computer calculation method to carry out the knowledge graph study and the reasoning,and realizes the chronic kidney disease traditional Chinese medicine diagnosis and treatment experience auxiliary inheritance.The main research work is as follows:1.According to the diagnosis and treatment process of chronic kidney disease,construct the body layer of TCM diagnosis and treatment,and retain the complete path of TCM diagnosis and treatment in medical records.For chronic kidney disease TCM medical case literature,aiming at the problem of ignoring local word information in text processing,this paper proposes a method of identifying named entities based on local information and context information.The F1 value of the algorithm is 97.39%.The recognition entity is embedded in the diagnosis and treatment ontology layer to construct the knowledge map of chronic kidney disease.2.To the problem that the existing translation model can not fit the Chinese medicine knowledge to carry on the effective representation study,we introduce the Chinese phonetic transcription,stroke and structure information as the Chinese character and encodes.We propose the characteristic character substring PSS with the unique Chinese character(Pinyin,Stroke,Struct).To solve the problem that the effect of translation model only depends on the training data set,we propose to adjust the training and test data set dynamically,and use the translation model for iterative triple representation learning.Experimental results show that the characteristic substring PSS and iterative training decrease Mean Rank(MR)by 5.88 and 10.63.3.According to the diagnosis and treatment process of chronic kidney disease,the reasoning task of TCM knowledge map is decomposed into the task of link prediction between multi-layer entities of knowledge graph.Based on the knowledge map representation learning of chronic kidney disease,the paper finds hidden features and solves the link prediction of 1-N relationship by multidimensional convolution of entities and relationship embeddings in the knowledge graph.The experimental results show that multiple evaluation indexes are over 95.00%.4.Based on the knowledge graph construction,learning and reasoning work,the prototype system of knowledge graph learning and reasoning of traditional Chinese medicine is constructed,and the data management of knowledge graph of traditional Chinese medicine is designed and realized,including the functions of data addition and deletion,visualization,document acquisition and analysis,automatic question and answer,user management and so on.