Advanced Triglyceride Prediction and Applying Machine Learning Techniques for Better Health Diagnostics

The level of triglycerides in the blood is a key indicator of cardiovascular disease, diabetes, kidney disease, thyroid disorders, and liver disease. This study explores advanced regression analysis and feature selection techniques to improve the assessment of triglyceride levels. Extreme dimensionality reduction methods, including Principal Component Analysis and Independent Component Analysis, along with Lasso regression for feature extraction, were employed to improve feature characterization and eliminate noise. Several machine learning models—CatBoost, XGBoost, Random Forest, and LightGBM—were evaluated using R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, RMSE, and MAE as performance metrics. Among them, CatBoost demonstrated superior performance due to its ability to capture nonlinearity and complex interactions. The study also highlights the impact of feature engineering, with Principal Component Analysis and Lasso improving model efficiency by reducing redundant features. In addition, ensemble techniques such as stacking and bagging improved model stability and predictive accuracy. Explainability methods, including SHAP and LIME, were used to identify key characteristics that influence triglyceride levels, providing valuable information on prediction characteristics. These findings emphasize the importance of robust models, effective feature selection and interpretability techniques in achieving reliable triglyceride level predictions, which ultimately supports clinical decision-making in healthcare.