Feature Importance and Correlation for Enhancing ML Performance in Predictive Healthcare Analytics

Cardiovascular disease (CVD) is a predominant cause of death worldwide, highlighting the critical necessity for effective early detection and accurate diagnostic systems. As healthcare technology advances, machine learning has emerged as a transformative tool to improve diagnostic precision. This study investigates the impact of feature correlation on the performance of machine learning models for heart disease prediction across diverse datasets: Heart Disease Dataset (1190 instances) and Cardiovascular Disease (42,500 instances). We evaluated nine algorithms: Logistic Regression (LR), Extreme Gradient Boosting (XGBoost), Random Forest (RF), Decision Tree (DT), Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neigh-bor(KNN), Gradient Boosting (GB), and Multilayer Perceptron (MLP) using performance metrics such as accuracy, precision, recall, and F1 score. The analysis reveals that models like XG-Boost, RF, and MLP consistently deliver superior performance in datasets where key features are strongly correlated with the target outcome. Conversely, datasets lacking such correlations, particularly the Cardiovascular Disease dataset, exhibit significantly lower accuracy and overall model performance. These findings emphasize the crucial role of feature importance and correlation in shaping model accuracy and robustness, indicating that predictive model effectiveness depends on the relevance and strength of the selected features.