Life-style data-based multiclass weight problems prediction with interpretable ensemble fashions incorporating SHAP and LIME evaluation

Obesity is a significant public well being concern. Predicting weight problems danger from way of life information can information focused interventions, however present fashions stay restricted. This research first evaluates ensemble studying strategies after which combines approaches to enhance prediction accuracy and generalizability. Four ensemble techniques-boosting, bagging, stacking, and voting-were examined. Five boosting and 5 bagging fashions have been constructed alongside voting and stacking fashions. Hyperparameter tuning optimized efficiency, and have significance evaluation guided potential characteristic elemination. In part two, hybrid stacking and voting fashions built-in the best-performing boosting and bagging fashions to boost predictive functionality. Model robustness was ensured by way of k-fold cross-validation and statistical validation. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) improved interpretability by analyzing characteristic contributions. Hybrid stacking and voting fashions outperformed different ensemble strategies, with stacking reaching the very best efficiency (accuracy: 96.88%, precision: 97.01%, and recall: 96.88%). Feature significance evaluation recognized key predictors, together with intercourse, weight, meals habits, and alcohol consumption. The outcomes demonstrated that hybrid ensembles considerably improved weight problems danger prediction whereas preserving interpretability. Integrating a number of ensemble and explainability strategies offers a dependable framework for weight problems prediction, supporting scientific choices and personalised healthcare methods to mitigate weight problems danger.