Statistical Models For Disease Diagnosis Based On Electronic Medical Record Data

In recent years,most hospitals have gradually transformed from digital hospitals to smart hospitals,and the integration of "Internet + medical and health care" has become closer.Research on assisted disease diagnosis,medical image recognition and clinical decision support has been successively published,and medical services are gradually becoming intelligent.At present,there are still challenges in the application of "Internet + Medical" technology in medical clinical decision making.First,it is still a challenging problem that has not been truly solved to automatically extract knowledge from the massive electronic medical record database and establish a computable knowledge project.Secondly,most of the existing diagnostic statistical models of disease are based on structured data.In a word,it is still one of the hot issues in the medical industry how to make disease diagnosis and prediction according to patients’ description of their own conditions through learning from massive electronic medical record data.In order to solve the above challenges,this paper starts from the machine learning classification algorithm and establishes a statistical model of disease diagnosis through the learning of mass text electronic medical records,so as to assist doctors with insufficient clinical experience to make rapid and accurate clinical diagnosis.The main research contents of this paper include:1.Structured processing of 15,304 electronic medical records from a third-class A hospital in Hunan Province by using natural language processing technology to improve the data quality of the research.2.Naive Bayes,K-nearest neighbor and support vector machine methods were used to explore the establishment of diabetes diagnosis model,and confusion matrix was used to compare and analyze the model.The research results of this paper are as follows: on the one hand,natural language processing technology is used to clean and structure the original electronic medical record data,and the obtained structured data can be used for modeling research;,on the other hand,based on naive Bayes,K nearest neighbor,and support vector machine(SVM)method to construct the diagnosis of diabetes model,through theory and practice of authentication,the auxiliary diagnostic accuracy high,means that these three diseases diagnosis model can enter the hospital information system and use,can assist clinicians in the diagnosis and treatment of patients with diabetes,To assist clinical decision making effectively.