Cardiovascular Disease Risk Prediction Based on Random Forest

摘要

Cardiovascular disease (CVD) has caused great harm to human life and health and is one of the most serious diseases in the word. Traditional CVD prediction model to use more rigorous mathematical model approach, the inclusion factor, and the data format has high requirements. However, some medical data have a large number of missing attribute values, and these methods can not be fully adapted. To solve this problem, a CVD prediction model using machine learning algorithm is proposed. By further analyzing the feature contribution of the sample data, NB, SVM, DT, LR, RBF, and RF models are, respectively, constructed to predict cardiovascular diseases. Experimental results show that with the increase in data set capacity, even if it contains a lot of missing data, the effectiveness and capabilities of the proposed RF algorithm on the 2-category data set is superior to the above other machine learning algorithms mentioned. It is sensitivity is 88.0%, specificity is 87.6%, precision is 88.0%, and AUC value is 94.7%, respectively.