This study synthesized a novel and efficient magnetic adsorbent from activated carbon (AC) produced from waste sunflower seed shells (SSS). The synthesis steps included initial preparation of the shells, carbonization, chemical activation with various agents (KOH, Na₂CO₃, and H₃PO₄), and finally magnetic modification. FTIR, XRD, FESEM, EDX, and BET characterizations confirmed the successful combination of adsorbents. The experimental param eters were systematically designed and optimized using Response Surface Methodology (RSM) to maximize methylene blue (MB) adsorption. The quadratic model identified the optimal conditions for MB removal as follows: a contact time of 90 min, an adsorbent dosage of 0.03 g, and a solution pH of 12, utilizing the KOH activator. The mechanism and behavior of the adsorption process were analyzed through kinetic, isotherm, and thermodynamic studies. The kinetic data were well-described by the pseudo-second-order model. Furthermore, the results indicated that the Langmuir isotherm model showed a better fit with the data, suggesting monolayer adsorption. The thermodynamic investigation confirmed the spontaneous and endothermic nature of the adsorption process. Among the evaluated machine learning (ML) models, XGBoost demonstrated superior pre dictive performance, achieving the highest accuracy (R2 = 0.918) in forecasting MB adsorption. Notably, among the different activating agents, the adsorbent activated with Na₂CO₃ exhibited the highest adsorption capacity, while the one activated with H₃PO₄ showed the lowest efficiency, despite likely having a higher surface area. These results imply that besides surface area, other factors such as surface chemical properties (e.g., functional groups, surface charge, and polarity) play a determining role in the adsorption mechanism
