Prediction Of Cardio-cerebral Vascular Diseases Based On Electronic Medical Record Mining

With the rapid development of Internet technology,medical informatization and intelligentization has become a mainstream development trend.As an important carrier of medical information,electronic medical record(EMR)solves the problems of traditional paper-based medical record writing confusion,not easy to transmit and save,and becomes an important clinical data resource.At present,a large number of electronic medical record data have been accumulated in the major hospitals.How to use data mining technology to analyze and process the relevant medical record data and explore the potential value of medical record data is of great research significance.Taking cardio-cerebral vascular medical records as the research object,this thesis proposes two new disease prediction model aiming at the problems existing in the current research of EMR mining,such as insufficient utilization of medical record data,less risk factors to be considered,and single prediction results,and makes some progress in theory.The main research contents and achievements of this thesis are as follows:(1)A multi-view classification prediction model based on SVM is proposed.Through the analysis of EMR,it is known that not only structured medical record data but also text data features exist in the medical records.These text data features are also important factors affecting disease prediction.In order to make better use of the two aspects of data,this thesis uses a multi-view classification method for disease prediction.In addition,in order to reduce the influence of redundant and weakly correlated features in medical record data on the prediction model,the existing multiview classification model was improved,feature selection and prediction classification were unified into a learning paradigm,and a Multi-view Support Vector Machine Classification with Feature Selection(MSVMCFS)was proposed.The experiment takes the features of medical examination data and text record data as the two views to train the prediction model,and compared with other models,the results show that the model has the highest prediction accuracy and has good applicability.(2)A combined multi-label classification prediction model based on ensemble learning is proposed.Cardio-cerebral vascular disease is a multiple concurrent disease,especially for middle-aged and elderly patients,the patients have complications such as hypertension and diabetes while suffering from cerebral infarction.The existing disease prediction models mainly focus on single disease,but do not consider the complications of the disease.Therefore,this thesis adopts multi-label learning algorithm to jointly predict the disease.In order to improve the prediction accuracy and the stability of the model,a Bagging-based Combined Multi-label Classification(BCMC)algorithm is proposed to improve the existing multi-label model.First,the bagging algorithm is used for the initial integration of the three algorithms: the Classifier Chains method,the Label Powerset method,and the Rank SVM.On this basis,the algorithm results after the initial integration are voted to obtain the final prediction result.The experimental results show that compared with the traditional multi-label model,the multi-label prediction model through the secondary integration not only improves the prediction accuracy,but also the performance of the model is more stable.