Research On Cardiovascular Disease Prediction Based On Support Vector Machine And DS Evidence Theory

Cardiovascular disease is a very harmful disease to the body.In recent years,the incidence and death rate of cardiovascular diseases in China has been increasing year by year,which has brought serious negative impact on our society and economy.Cardiovascular disease is the leading cause of death in China and even in the world.Cardiovascular disease is characterized by high mortality and high disability,making it difficult for people suffering from cardiovascular diseases to heal.Therefore,primary prevention of cardiovascular disease is particularly important.Many studies have shown that most of the major risk factors for cardiovascular disease can be artificially adjusted.Therefore,it is necessary to develop accurate and efficient early prediction tools for cardiovascular diseases to identify high-risk groups and provide early warning of the disease.At the same time,it is suggested that high-risk groups change their unreasonable living habits to adjust risk factors,thereby reducing the risk of cardiovascular disease.In this paper,based on the analysis of existing disease prediction methods,in view of the advantages of SVM algorithm and DS evidence theory,a SVM-DS disease prediction model is proposed for cardiovascular disease prediction.The model is based on the SVM algorithm to model,use the established model to predict the disease condition,convert the posterior probability of the SVM model output to BPA function,introduce the DS evidence theory,synthesize the BPA function,and output the final prediction result according to the decision rule.Next,it proposes to divide the disease-related risk factors into major risk factors and potential risk factors,and analyze the two risk factors separately.Take heart disease as an example,and carry out logistic regression multifactor analysis on risk factors related to heart disease.There were risk factors that were significantly associated with heart disease.Then,the data set was randomly sampled at a ratio of 2:1:1 and divided into a training set,a validation set,and a test set.Then the training set,verification set,and test set are divided into 6 corresponding data sets according to the criteria for dividing risk factors.Then,the disease prediction model was established based on Logistic regression,SVM and SVM-DS.Logistic regression,SVM was based on training set and validation set modeling,SVM-DS was based on training set modeling,and the model predicted TPR and TNR of validation set.As a basis for the construction of BPA functions.During the training process,the SVM model's penalty factor and kernel function parameters are debugged to find the optimal parameters of each model.Finally,the test results of the three models are tested on the test set.Three models have good predictive performance on the test set.The overall recognition rate of the SVM-DS model is the highest,and the recognition rate of positive and negative classes is higher.And found by calculating the area under the ROC curve AUC:0.8889_SVM?0.8899_Logistic?0.9052_SVM-DS,the SVM-DS model is the best performance model.Experiments show that for the prediction of heart disease,the prediction results of the SVM-DS model are more reliable and the performance of the algorithm is more stable.The SVM-DS model proposed in this paper has certain reference significance for the early warning of heart disease.The model can be applied to other cardiovascular disease predictions after being extended.