GRADIENT BOOSTING-BASED HEART FAILURE PREDICTION FOR IMPROVED MEDICAL DECISION MAKING

Abstract

Heart failure is a life-threatening cardiovascular disease that has a high mortality rate, where the heart is unable to pump blood effectively. Early and precise prediction is essential for timely intervention and better patient outcomes. Machine learning has demonstrated potential in discriminating high-risk patients using clinical data, yet most of the literature applies a single algorithm, which restricts predictive accuracy. The proposed study uses the state-of-the-art gradient boosting algorithms XGBoost and CatBoost to predict heart failure. The suggested methodology involves preprocessing that involves dealing with missing values, eliminating outliers, and normalization of features to guarantee the model with reliable input. Experimental evidence has shown that XGBoost is better than CatBoost with an accuracy of 77.78% and a lower false negative rate, showing better capability of identifying at-risk patients. Incorporating these models into the clinical decision support systems can lead to better early detection, timely treatment, and probably mortality associated with heart failure.