Influencing Factors and Prediction of Heart Disease

Abstract

Heart disease (HD) is a key issue of concern in the global safety and health field nowadays. Some researchers listed the predisposing factors and established some models to predict the incidence rate. However, there is still a gap in the research on the influence of related factors and the comparison of different models. Therefore, the research topic of this article is the importance of HD factors and the comparison of model predictions. Following are this article’s research methods: Firstly, this project preprocesses and visualizes the data based on the Heart Failure Prediction Data Set. Then this article establishes three mainstream research models, including Logistic Regression (LR), Random Forest (RF), and Support Vector Machine (SVM) model. Finally, this article introduces the Receiver Operating Characteristic (ROC) curve for better comparison of models and illustrates the importance of variables on HD through the ranking graph. Research has found that the RF model has the highest prediction accuracy, reaching 86.34%, and its Area Under Curve (AUC) is also the largest. Therefore, through comparison, the predicted accuracy of the RF model is the best. Meanwhile, in the ranking image of variables, it can be concluded that the Slope of peak exercise ST segment (ST_Slope) has the greatest impact on HD. Due to the possibility of prediction errors and limited applicability, it is possible to consider improving the variables to enhance model accuracy and contribute to health.