Towards Precision Public Health: Advancing Obesity Prediction with Machine Learning Algorithms

Addressing the global epidemic of obesity requires innovative approaches that enhance predictive accuracy and provide actionable insights. This study introduces a groundbreaking application of machine learning (ML) algorithms, enriched with SHapley Additive exPlanations (SHAP), to predict obesity with unprecedented precision. By deploying a comprehensive suite of algorithms—CatBoost, Gradient Boosting, XGBoost, LightGBM, Random Forest, Bagging, and a novel Stacking Hybrid model—this research systematically quantifies and compares their predictive performances on a multifaceted dataset representing diverse demographic attributes. The incorporation of SHAP values into these algorithms marks a significant novelty, facilitating a nuanced understanding of the factors influencing obesity predictions. This approach not only enhances model transparency but also contributes to the robustness of obesity prediction by revealing the impact of each feature on the model’s output. Our findings indicate that the Stacking Hybrid model achieves the highest accuracy, with a balanced accuracy score significantly outperforming conventional models. This research advances the frontier of predictive analytics in public health by demonstrating the efficacy of ML algorithms augmented with explainability in tackling complex health challenges such as obesity. The superior accuracy and interpretative depth offered by our approach hold profound implications for policy-making and health interventions, paving the way for precision public health practices that are both scientifically sound and ethically responsible.