A Boosted Ensemble Framework with Bio-Inspired Optimization for Multivariate Classification of Structured Data

This study presents a comprehensive machine learning framework that integrates advanced ensemble classifiers with bio-inspired optimization to perform high-accuracy classification of multivariate structured data. The proposed approach is applied to classify countries based on health, demographic, and economic indicators. Specifically, three state-of-the-art boosting models, Extreme Gradient Boosting (XGBoost), LightGBM, and Histogram-Based Gradient Boosting, are evaluated for their predictive performance. To improve generalization and accuracy, the Artificial Protozoa Optimization (APO) algorithm is employed for hyperparameter tuning. The resulting hybrid model, XGAO (XGBoost [Formula: see text] APO), achieved the highest classification accuracy of 0.988. A fast sensitivity analysis-based feature selection method was used to reduce dimensionality and enhance interpretability. Key features such as Adult Mortality, Diphtheria, and Population emerged as strong predictors. SHAP (SHapley Additive Explanations) analysis was conducted to provide insight into feature contributions, supporting transparency and model understanding. The study also visualized feature relationships using a correlation plot with clean labels to emphasize data distributions and interdependencies. Compared with baseline models and other recent techniques, the proposed framework demonstrated superior classification accuracy and robustness. The results validate the integration of gradient boosting and metaheuristic optimization as an effective strategy for interpretable and scalable classification on real-world tabular datasets. These contributions are particularly relevant to applications in socioeconomic analytics, intelligent decision support systems, and uncertainty modeling.