Machine learning for optimizing Reactive Red 120 biosorption onto cross-linked gelatin/alginate-TTAB-montmorillonite...

A biodegradable thin-film composite comprising cross-linked gelatin/alginate and tetradecyltrimethylammonium bromide-modified montmorillonite was developed for the adsorption of Reactive Red 120 (RR120). Compared with our previously reported bulk composite, the thin-film architecture shortened diffusion pathways, thereby reducing internal mass-transfer resistance and improving adsorption-site accessibility. Batch adsorption experiments showed that the equilibrium adsorption capacity (q_e) increased with contact time, initial dye concentration, and temperature, but decreased with increasing solution pH. Kinetic modeling using the pseudo-first-order, pseudo-second-order, and intraparticle diffusion models indicated that adsorption was governed by surface adsorption and intraparticle diffusion, whereas equilibrium data were best described by the Langmuir model with a maximum adsorption capacity (q_m) of 46.79 mg/g. Regeneration using 0.5 M NaCl demonstrated moderate reusability. Among the evaluated machine-learning algorithms, the optimized gradient boosting model achieved the highest predictive performance (test R² = 0.9606, RMSE = 1.4265 mg/g) and predicted optimum operating conditions of 356 min, 277 mg/L, pH 2.8, and 41 °C. Experimental validation yielded an adsorption capacity of 45.23 ± 0.05 mg/g, with a relative prediction error of only 0.91%. These findings demonstrate that integrating biodegradable thin-film composites with established machine-learning techniques provides an effective strategy for adsorption-process optimization and textile wastewater treatment.