Akademik Veri Yönetim Sistemi
WATER AIR AND SOIL POLLUTION, cilt.237, sa.6, 2026 (SCI-Expanded, Scopus)
This study aims to provide a sustainable solution for dye removal by developing environmentally friendly and renewable resource-based adsorbents. In this context, the removal of Rhodamin B (RhB) dye from aqueous solutions was investigated using composite biochar obtained from the seaweed species Ulva lactuca -Cystoseira sensu lato. The characterization of the adsorbent, which draws attention with its high surface area (797.91 m2/g), porous structure and rich functional groups, was carried out by SEM, FTIR and BET analyses. The data obtained as a result of adsorption experiments were evaluated with kinetic and isotherm models; the pseudo-second-order kinetic model (R2 = 0.993) and Freundlich isotherm (R2 = 0.994) successfully reflected the chemical and multi-layered nature of the process. In this context, the adsorption data were modelled with four different machine learning algorithms (Random Forest, XGBoost, CatBoost, Gradient Boosting). Among the models, Random Forest algorithm showed the highest accuracy on the test data (R2 = 0.93, RMSE = 4.15, MAE = 1.95) and successfully reflected the behaviour of the system. The decision mechanisms of the model were explained with SHAP analysis; the dominant effects of contact time and adsorption capacity on the prediction were determined. While PDP analyses visualized the individual effects of variables on the target variable, learning curves and residual analyses confirmed the generalization ability and stability of the Random Forest model. This multi-faceted approach shows that both biomass wastes can be converted into environmentally friendly adsorbents and machine learning-based methods can be used effectively in water treatment applications.