Evaluation of Feature Attributes for the Prediction of Heart Disease Using Machine Learning

<ns3:p> Background Every year, approximately 20.5 million people die of cardiovascular diseases (CVDs). Early detection of CVD helps people to treat it. As a result, patients can alter their daily schedules and, if required, take medications. According to World Health Organization (WHO) reports, CVD causes approximately 20.5 million people annually. By 2030, these deaths is expected to reach 24 million, accounting for 31.5% of all deaths worldwide. According to a WHO study, medication therapy and patient counseling are also necessary to lower the risk of heart attack and stroke by 2025. <ns3:sup>1,2</ns3:sup> Methods For the early prediction of CVD, six machine learning methods, including the regression model, naïve Bayes, random forest, logistic regression, XGBoost, and LightGBM, were employed. Thirteen features were chosen for training. The models were trained in three ways, namely with full thirteen features, features selected by the chi-square test, and features with 0.75, 0.5 correlated values between each other. The performance metrics considered for the evaluation of the model were accuracy, F1-score, recall, and precision. Results Random forest provided 99% of the highest accuracy by considering all features. Feature reduction based on correlation was used for training, and accuracy was evaluated. Python scripting language was employed to implement the proposed model. </ns3:p>