Visible-light photocatalytic activity of a green niobium pentoxide nanocatalyst: Experimental analysis and machine learning approach

This work aims to develop green Nb₂O₅-NPs from Carya illinoinensis extract for the photodegradation of six organic dyes and to predict the reaction degradation progress through a machine learning study. 71.5 % removal was the highest value achieved for dye degradation after 120 min under visible radiation with the apparent rate of the pseudo first-order reaction k = 0.0034 min⁻¹. DT algorithm showed satisfactory performance (R²_training = 0.9975 / R²_test = 0.9945, RMSE_training = 0.059 / RMSE_test = 0.094), predicting 75 % of MB dye removal after 300 min under visible radiation. Nb₂O₅-NPs showed high phytotoxicity for Beta vulgaris L. seeds (except at 50 mg L⁻¹) and no phytotoxicity for Brassica oleracea seeds in all concentrations tested (12.5–100 mg L⁻¹). Therefore, this study highlights the importance of integrating machine learning models into catalytic research, offering valuable insights for optimizing reaction conditions and guiding scalable photodegradation processes.