Artificial Intelligence-Assisted Prediction and Optimization of Methylene Blue Adsorption onto Albizia lebbeck Bio-adsorbent Using Ensemble Machine Learning

This study investigates the adsorption of methylene blue (MB)from aqueous solution using a bio-adsorbent prepared from Albizialebbeck pods and evaluates the predictive performance of ensemble machine learning models for modeling the adsorption process. The bio-adsorbent was characterized using pH PZC , Fourier Transform Infrared spectroscopy (FTIR), X-ray Diffraction (XRD),and Scanning Electron Microscopy with Energy Dispersive Spectroscopy (SEM–EDS). Characterization revealed a porous, carbon-rich, predominantly amorphous structure with abundant oxygen-containing functional groups favourable for dye adsorption. The effects of solution pH, contact time, adsorbent dosage, and initial dye concentration were systematically investigated through batch adsorption experiments. Equilibrium and kinetic data were analysed using Langmuir, Freundlich, and Temkin isotherm models, as well as pseudo-first-order, pseudo-second-order, and intraparticle diffusion kinetic models. The Langmuir isotherm provided the best equilibrium fit, yielding a maximum monolayer adsorptioncapacityof424.24mgg − 1 , while the pseudo-second-order model best described the adsorption kinetics. Furthermore, ensemble machine learning models—Random Forest (RF) and eXtreme Gradient Boosting(XGBoost)—were developed and optimized to predict adsorption performance. The optimized RF model achieved an R 2 value of 0.98876 for MB removal efficiency, while XGBoost attained an R 2 of0.99953 for adsorption capacity prediction. These findings demonstrate that Albizia lebbeck bio-adsorbent is a highly efficient, eco-friendly material for MB removal, and that ensemble learning provides a robust and accurate framework for predicting and optimizing adsorption processes in wastewater treatment.