Akademik Veri Yönetim Sistemi
PROCESS SAFETY AND ENVIRONMENTAL PROTECTION, ss.1-33, 2026 (SCI-Expanded, Scopus)
Rapid urbanization has intensified heavy metal contamination in children's playgrounds, posing significant risks to both urban sustainability and children's health. This study assessed heavy metal contamination in surface dust at children’s playgrounds in Ankara (Türkiye) by combining experimental analysis with machine learning methods. The study employed a prediction framework based on the XGBoost model, which includes six different meta-heuristics: Artificial Bee Colony (ABC), Bat Algorithm (BA), Dwarf Mongoose Optimization Algorithm (DMOA), Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO), and Whale Optimization Algorithm (WOA). This framework measures cumulative non-cancer health risks for children using the Hazard Index (HI). Dust samples were gathered from fifty sampling points (n = 50), and heavy metal concentrations were measured using ICP-OES; HI values were calculated by considering exposure pathways specific to children. Each optimization algorithm was run separately 10 times without setting a fixed random seed, and the performance metrics are reported as the mean and standard deviation. The results indicated that the XGBoost model optimized with the Whale Optimization Algorithm (WOA-XGBoost) achieved the highest performance; it reached an R² value of 0.869 on the test set and the lowest error metrics (MSE = 0.005; RMSE = 0.064; MAE = 0.050; MAPE = 5.205). Furthermore, SHapley Additive exPlanations (SHAP) and Individual Conditional Expectation (ICE) analyses clarified the model’s decision mechanism, and identified the key metals influencing HI along with their threshold behaviours. The findings contribute to sustainable urban development, inform urban environmental quality management, and provide a scientific basis for safeguarding children's health.