Performance of Ensemble Learning Models for Non-Invasive Total Cholesterol Level Estimation

Non-invasive estimation of cholesterol levels has emerged as a crucial area in healthcare, offering a comfortable alternative to traditional invasive methods. This study leverages machine learning techniques to predict total cholesterol levels using readily available demographic and vital data. The dataset encompasses 390 data points and 16 attributes, including age, gender, and blood pressure measurements. We explore the potential of ensemble models, such as Random Forest, Gradient Boosting, AdaBoost, and Stacking to enhance prediction accuracy. The Stacking Regressor emerges as the standout performer, achieving an MAE of 17.61 and an RMSE of 21.97. This model significantly outperforms previous research in terms of accuracy as its RMSE is better by 36% in comparison with the previous comparative methods. Also, it doesn't require complex and intrusive data collection methods. Our research bridges the gap between machine learning and non-invasive diagnostics, signaling promising advancements in healthcare diagnostics.