A hybrid RSM-ANN-NSGA-II framework for optimizing plant-mediated green synthesis of silver nanoparticles from coriander stem extract

Plant-mediated green synthesis of silver nanoparticles (AgNPs) offers a low-toxicity route to antibacterial nanomaterials, but the strong non-linear coupling between synthesis variables makes it difficult to identify favourable operating conditions from a limited number of experimental runs. This work presents a hybrid RSM-ANN-NSGA-II modelling and optimization framework and applies it to the green synthesis of AgNPs from coriander ( Coriandrum sativum ) stem extract. A four-factor Box-Behnken Design (BBD) was used to study the effect of coriander extract concentration (CE), AgNO 3 concentration, reaction temperature and volume ratio on two responses: particle size (nm) and antibacterial zone of inhibition (ZOI, mm) against Staphylococcus aureus . Six data-driven models – a quadratic response surface model (RSM), an artificial neural network (ANN), Random Forest, Gradient Boosting, Support Vector Regression (SVR) and Gaussian Process Regression (GPR) were fitted to the 27-run BBD data and compared under leave-one-out (LOO) and 20-seed 5-fold cross-validation. The quadratic RSM was the most accurate and most reproducible model (LOO R 2 = 0.944 for particle size and 0.933 for ZOI), while the ANN showed strong overfitting behaviour typical of small-data regimes (LOO R 2 collapsing to strongly negative values). The RSM was therefore adopted as the primary surrogate for a multi-objective optimization with the non-dominated sorting genetic algorithm II (NSGA-II), which returned a Pareto front of solutions trading off minimum particle size against maximum ZOI. NSGA-II converged reproducibly across 20 independent random seeds. Formation of the AgNPs was confirmed by a color change from pale yellow to brown, a characteristic surface plasmon resonance peak near 390 nm in the UV–vis spectrum, an average hydrodynamic size of 74.6 nm from dynamic light scattering, and a zeta potential of −5.54 mV. Transmission electron microscopy revealed discrete AgNP cores in the ∼40–55 nm range, and scanning electron microscopy showed the aggregated dried morphology consistent with the moderate zeta potential. The results confirms that the hybrid RSM-ANN-NSGA-II framework is a practical tool for guiding plant-mediated AgNP synthesis with limited experimental effort, provided that surrogate choice is validated by proper cross-validation rather than by in-sample training accuracy.