Deep Learning Algorithms For Predicting Cardiovascular Disease

Prediction of cardiovascular diseases is a major challenge as they are the leading cause of death worldwide.Many scholars have used different methods such as mathematical models and machine learning to predict the probability of occurrence of the disease,trends and various other scenarios.However,not much research has been done to predict cardiovascular diseases.In order to effectively prevent cardiovascular diseases and enable doctors to diagnose them in a timely and effective manner,this study uses LSTM algorithm to build a cardiovascular disease prediction model and verify its feasibility and performance.Based on the analysis of domestic and international literature,it can be found that among the traditional machine learning methods for disease prediction,the support vector machine(SVM)model is one of the most widely used models,and the accuracy of the support vector machine is relatively high in terms of the final prediction results of the models in the literature.Also in predicting cardiovascular diseases many researchers have used the K-nearest neighbor(KNN)algorithm to build the model.Therefore,the study uses both support vector machine and K-nearest neighbor models in the performance validation work of LSTM prediction models.The main research of the article is as follows:(1)Obtain the feature attributes needed to build the model in the UCI database and Kaggle platform.The 87 feature attributes extracted were subjected to culling and ranking operations using random forest,while 27 feature factors with strong correlations were finally extracted based on the existing medical literature and the Fereham table.The cardiovascular dataset was subjected to data processing techniques to obtain 1024 case data for model training and testing.(2)After screening out the dataset and feature attributes the associated LSTM model structure needs to be built.The accuracy of the model is compared for different parameters.The model is trained using the delineated dataset and the parameters are continuously optimized using an optimizer to build the final LSTM model.The test set is used to analyze the effectiveness of the model,and the created LSTM model is evaluated using four metrics: Precision,Recall,Accuracy,and AUC.In addition to verify the feasibility of this LSTM model,it was compared with studies that also used the LSTM algorithm to predict diseases.(3)The SVM and KNN prediction models used for the comparison were created.After selecting the appropriate model parameters,both models were built with the same dataset and evaluated for performance using the same four metrics(precision,recall,accuracy,and AUC values),respectively.Finally,when the evaluated metric values of these three models were compared,the data showed that the LSTM model had higher values for each metric than the other two models,so it was concluded that the LSTM prediction model was the best for this application of predicting cardiovascular disease.