Research And Application Of Chronic Disease State Prediction

After the epidemic of coronavirus pneumonia,more and more people realize the importance of basic medical and health care capabilities.The application of artificial intelligence technology in the process of disease prevention and treatment can deeply discover the importance of science and technology for the protection of people’s life safety.With the progress on cloud computing,massive medical data is gradually stored in the cloud.People intend to make use of machine learning algorithm to mine and analyze,in order to achieve the purpose of disease prediction and diagnosis.At present,the medical industry is more and more willing to adopt machine machine learning methods for research,because it is conductive to improve the scientificity and rationality in the process of medical diagnosis and analysis.Establish a reasonable disease prediction and diagnosis model,analyze the association and distribution rule hidden in the medical data,so as to continuously learn the medical data in different stages.At the same time,the research on the diagnosis and treatment service based on clinical medical data can further enrich the theoretical research system of online medical service,and expand the application prospect of online medical service,which has important clinical significance.Therefore,the essay studied the prediction methods of chronic disease state,in order to provide a more intuitive and powerful scientific tool for the intervention of patients’ behavior.Firstly,the essay investigated the research stage of medical disease prediction in the world,introduced the data conversion and feature crossover methods in data mining theory,and then introduced commonly used prediction models and model fusion theory,focusing on the working principle of the algorithms.The essay has made use of the dataset from UCI open-source database called Infarction complications.Specifically,the dataset describes the patient information of routine examination in hospital and final recovery status.Becasuse of the imbalance distribution of four kinds of recovery status,the essay proposed the NR-LGBM discharge recovery state prediction model.After processed by the noise reduction autoencoder,the dataset would go through binary classification experiment at first,in order to select sample without any complications.Then the remaining sample data would be processed by multiple classification experiment.After comparing the performance of KNN and LDA classification models,the NR-LGBM model has shown a better classification performance with Macro-F1 value of 0.811.The essay also made use of the dataset called Heart_disease from UCI to carry out the model fusion experiment.Three independent models including SVM,decision tree and logistic regression have been used to ensembled constructing the Eclf heart disease prediction model.Compared with three individual models,the Eclf fusion model has shown the best classification performance with F1-score of 0.83.As a result,the essay has explored efficient solutions for the application scenarios of discharge recovery state prediction and heart disease prediction,which has clinical significance and application prospects in the field of medical services.