Research On Core Herb Discovery And Prescription Recommendation Algorithm Based On TCM Prescription Knowledge Graph

The intelligentization of medical diagnosis and treatment is the trend of future development,while traditional Chinese medicine is an important option to solve the dilemma of modern medical treatment.So the research on informatization of traditional Chinese medicine has attracted the attention of many scholars,and the construction of the knowledge graph of traditional Chinese medicine is a research hotspot.The development of in-depth learning technology makes the knowledge graph of traditional Chinese medicine more and more complex and accurate,However,how to use TCM knowledge graph to better mine the knowledge hidden in the huge TCM prescription database and provide support for auxiliary diagnosis and treatment needs further research.This thesis completes the following four tasks for semi-structured data of traditional Chinese medicine prescriptions:(1)Build a knowledge graph with TCM prescriptions as the core.Compile a crawler program to crawl the regular semi-structured original prescription data and herb information from the Internet,and extract different types of entities such as prescriptions,herbs,symptoms and their relationships from the original data using the method of regular expression matching keywords and formats,and store the entities and relationships in the form of triplets into the Neo4 j database to build a knowledge graph of traditional Chinese medicine prescriptions.(2)Design a core herb discovery algorithm based on TCM prescription knowledge graph.First,use the knowledge graph to construct a heterogeneous information network containing different types of objects such as prescriptions,herbs,symptoms,functions,and select multiple meta-paths based on Path Sim algorithm to calculate the similarity between prescriptions,then cluster the prescriptions based on this.Finally,a calculation method of herb importance coefficient combining the TF-IDF algorithm principle and dose factor is designed and used to obtain the core herbs in each prescription cluster.(3)Design a prescription recommendation algorithm based on TCM prescription knowledge graph.First,use the knowledge graph to construct a two-type network composed of symptoms,prescriptions and their relationships,and use the Rank Clus algorithm to get the clusters and the ranking of prescriptions and symptoms on each cluster with the prescription as the target object.Then,based on this result and combining the diagnostic concept of traditional Chinese medicine in reality,design the subsequent prescription recommendation process to complete the entire prescription recommendation algorithm.(4)Design and implement an application platform of TCM prescription knowledge graph.Based on the completed knowledge graph of traditional Chinese medicine prescriptions,a knowledge graph application platform is built by using the Spring Boot framework to integrate the Neo4 j database,providing information retrieval and graphical display of knowledge graph and other functions,and the prescription recommendation algorithm designed in this thesis is further applied to realize the prescription recommendation function,which is convenient for users to use.