Prediction Of Cardiovascular And Cerebrovascular Diseases Based On Deep Learning

Cardio-cerebrovascular diseases have become the number one killer of human life and health.There are many factors that cause cardio-cerebrovascular diseases,and doctors will diagnose them based on various detailed medical reports.With the rise of artificial intelligence,deep learning has been applied to various fields and has a good effect.In the medical field,deep learning is also used to predict and judge the risk of disease,but its prediction accuracy when processing complex data still needs to be improved.This article mainly studies the prediction of cardiovascular and cerebrovascular diseases,uses improved genetic algorithms to extract feature attribute data from the original sample data,and uses the feature attribute data for deep neural networks and predicts cardiovascular and cerebrovascular diseases.The specific work includes two aspects:(1)Use improved genetic algorithm to select the characteristic attributes of sample data.In view of the problems of genetic algorithm(GA),such as local convergence,low operating efficiency,and low population fitness,this paper uses a combined algorithm based on CFS and genetic algorithm(CFS-GA),and CFS-EGA is formed by improving the elite selection part of genetic algorithm.The specific improvement method is to use the selection method of preserving superiority to eliminate inferiority,save excellent individuals,avoid the occurrence of degradation,and thus accelerate the speed of genetic evolution;improve genetics by introducing adaptive crossover,mutation operations,and the idea of acquaintance in crossover operations The global search capability of the algorithm.Finally,an experimental comparison is made on the selection of feature attributes of GA,CFS-GA and CFS-EGA.The results show that the CFS-EGA algorithm has better ability to select feature attributes than GA and CFS-GA.(2)Propose an improved neural network DBN-LSTM that combines a deep belief network(DBN)and a long-term memory network(LSTM).Use CFS-EGA to process the subset of feature attributes for training.In the process of training DBN-LSTM,perform the optimal selection test of the number of hidden layers on the upper DBN.At the same time,using the Taguchi method to select the lower LSTM hyperparameters,combined with the verification set to determine the LSTM hyperparameters.Construction of deep neural network(DNN),convolutional neural network(CNN),long and short-term memory network(LSTM),used for comparative analysis with DBN-LSTM.The classification method was used to compare the average of the final results of the two experiments.The results showed that the prediction accuracy of DBN-LSTM cardiovascular and cerebrovascular diseases reached 95.61%,which was higher than that of the three traditional neural networks.Finally,the regression method is used to compare the prediction capabilities and stability of DBN-LSTM and LSTM when faced with a different number of feature attribute subsets,which proves that DBN-LSTM has better performance and stability than LSTM.