Study of machine learning on the photocatalytic activity of a novel nanozeolite for the application in the Rhodamine B dye degradation

Abstract

Contamination of wastewater with organic dyes has caused a serious threat to humans and aquatic life due to the hazardous effect of these contaminants. In this context, the present work aims to carry out a Machine Learning (ML) study to evaluate the photocatalytic activity of a nanozeolite (nANA) in the degradation of Rhodamine B (RhB) dye. Three machine learning algorithms (Random Forest, Artificial Neural Network and Xtreme Gradient Boosting) were used in the regression model development. The dataset used in the machine learning and data correlation was generated by Central Composite Rotational Design (CCRD 2²). Regarding the machine learning study, the ANN with structure 3:6:1 showed the best performance as a predictive model (R² = 0.98 and 0.9 for training and testing, RMSE < 5.0), resulting in the 50.37 ± 1.01 % RhB removal at pH 5.7, [RhB] = 200 mg L⁻¹ and [nANA] = 2.75 g L⁻¹ after 180 min under visible light. Feature importance revealed that all parameters (pH, [RhB], [nANA]) were relevant to the response. Therefore, this work confirms the potentiality of machine learning algorithms to develop predictive models as well as a good starting point for the scale-up of advanced oxidation processes.