Research On The Method Of Doctor-patient Matching Based On Internet Medical Big Data

Internet medical treatment has been accepted and used by more and more people in recent years.Internet medical treatment has optimized the medical treatment process,alleviated the uneven distribution of resources and asymmetric information in the medical treatment.The internet medical platform has accumulated a large number of doctors' information,patients' information and patients' Q & A data,forming a valuable internet medical big data.In the internet medical model,one of the most mainstream service models is online consultation.When patients receive online consultation,they hope to find the most experienced doctors for their own diseases,and at the same time,doctors also hope that the diseases of the patients they consult are the direction they are good at.The problem of doctor-patient matching is universal and urgent.The mode of doctor-patient matching can not only reduce the waste of diagnosis and treatment resources,but also improve patient satisfaction.This thesis digs the internet medical big data accumulated in online consultation websites,and divides the task of doctor-patient matching into two steps: disease diagnosis and doctor recommendation.Disease diagnosis refers to: if the patients are not clear about their own diseases when they visit online,the improved disease diagnosis method combining knowledge graph and deep learning in this thesis can allow the patient to make a preliminary diagnosis of his own disease.This thesis introduced the BERT + Bi LSTM+CRF with Pinyin medical entity recognition model to more accurately recognize the medical entity in the patient's condition description.And the entity linking method based on multi domain index(IRNorm entity linking model)can link the recognized medical entity more accurately to the standard entity in the medical knowledge graph,so as to eliminate the adverse impact of the patients' irregular behaviors such as short description and wrong description of the medical entity on the accuracy of disease diagnosis.The experimental results also show that the improvement of the disease diagnosis method combining knowledge graph anddeep learning can effectively improve the accuracy of disease diagnosis.Doctor recommendation refers to that after the patients have recognized their own diseases,they need to find a good doctor for online treatment.In this thesis,combined with the features of doctors' social capital,online reputation and so on,the doctor recommendation problem is transformed into a sort learning problem of doctors.Through Lambda MART sort learning algorithm,we can get the comprehensive score and rank of doctors with various features of doctors,so as to realize doctors' recommendation on patients' diseases.The innovation of this thesis is mainly reflected in the following three aspects.The thesis improves the model of disease diagnosis based on knowledge graph and deep learning.The ranking learning is applied to doctors' recommendation,and the ranking is based on the features of multiple aspects of doctors.It enriches the research perspective and method in the literature of doctor-patient matching.