Enhancing Heart Disease Prediction with Multiple Imputation and Feature Selection in XGBoost

For cardiac disease prediction, this paper examines feature selection approaches such Particle Swarm Optimization (PSO) and the Chi-square test with XGBoost. We analyze their performance indicators to see how these two techniques affect predictive modeling. We employ Multiple Imputation by Chained Equations (MICE) for missing values and SMOTE for class imbalance to improve our model’s reliability. PSO-XGBoost and Chi-square-XGBoost predict similarly, with just minor differences in a few performance indicators. For optimal cardiac disease prediction, specialized feature selection procedures are essential, as shown by the comparison and Chi-square-XGBoost gives better accuracy with 92.2 %. To construct reliable cardiac disease prediction models, one must understand feature selection and preprocessing approaches.