Research On Knowledge Graph Medical Diagnosis Method Based On Deep Learning

Medical and health issues concern people’s physical health,life safety and family happiness,and protecting the physical and mental health of the people is the top priority.Nowadays,the pace of life of contemporary young people is gradually speeding up,the material needs are gradually improving,the pressure of life is increasing,and the chance of sudden death of young people is getting higher and higher.How to use wisdom medical technology,to achieve efficient and convenient way of online medical visits help ordinary users without leaving home to master medical knowledge,self-diagnosis and treatment under the emergency state,and through the popularization of medical knowledge to strengthen disease awareness and prevent the potential threat of chronic disease,to implement a regular check-up is the top priority of their bodies.In this context,this thesis studies disease diagnosis methods with the help of deep learning,knowledge graph and other technical means.The main research contents include the following three aspects:First,a vocab-GCN based Chinese medical text classification method is proposed,which can identify 45 different diseases or departments.The graph convolutional neural network model can directly learn the medical text graph,and save the global structural information of the medical relation graph in the graph embedding.The proposed method was compared with LSTM,GRU,CNN and GCN baseline model in the text classification task from the accuracy,precision,recall rate and F1 score.The experimental results show that vocab-GCN has better classification performance.The accuracy of disease classification reached 75.86%,which was 6.17% higher than the optimal deep learning method.Second,the medical diagnosis method based on knowledge graph is studied.The method first automatically builds a knowledge graph by extracting entities such as diseases,symptoms,and drugs in structured datasets and the relationships between entities.Secondly,the AC multi-pattern matching algorithm and the semantic similarity calculation method based on the three methods of edit distance,overlap coefficient and cosine similarity score are used to jointly identify entity information,and deeply analyze the input problem.Thirdly,the problem classification is converted into text classification through the Naive Bayes algorithm,and the optimal accuracy can reach 96.86%.Finally,the quality of knowledge graph construction was evaluated through the Q&A effect demonstration.After several tests,the Q&A system can answer most medical questions accurately.Third,a novel deep learning based knowledge graph medical diagnosis method is studied.In this method,an intelligent question answering matching method(BBBMU)based on BERT combined with Bi LSTM and Bi GRU is proposed,BBBMU and knowledge graph are used to solve medical diagnosis problems.For the input text question,the medical diagnosis method firstly performs information retrieval in the knowledge graph.If no answer is found,it is judged that the search intent is not clear,and the question pair retrieval module is entered.In the experiment,the proposed method was compared with the most advanced three disease diagnosis models of MALSTM,BERT and HBAM.The experimental results show that the disease diagnosis accuracy rate of this method is86.5%,and it has better medical diagnosis performance.