Adsorption of organic dyes onto nanozeolites: A machine learning study

Wastewater pollution with organic dyes has generated great concern in society due to the hazardous effects these contaminants pose to humans and aquatic life. In this view, the application of the adsorption process using nanoadsorbents has been a promising alternative due to the relatively low cost, high efficiency and simple operation. In addition, nanozeolites were highlighted in scientific literature due to their properties (surface area, porosity, ion exchange capacity, chemical and thermal stability), being useful for dye removal from wastewater. However, time and cost in experimental procedures are required to find optimal conditions for the adsorption of dyes onto these nanozeolites. Therefore, machine learning methods have emerged as a suitable tool for the prediction of the adsorption capacity of the nanoadsorbents in an efficient manner, being capable of recognizing patterns in the process and addressing the process feasibility. In this context, the present work aims to develop a machine learning (ML) study of the adsorption of organic dye onto nanozeolites and to identify the main vari-ables that affect the adsorption capacity and removal of organic dye from wastewater. Thus, four ML algorithms (RF, LGB, XBG, and ANN) were tested as a regression model. This study revealed that XBG showed the best performance in comparison to the other models, being suitable in the prediction of adsorption capacities of nanozeolites for cationic dyes. Additionally, an exploratory analysis and hypothesis testing confirmed the great effect of the dye and nanoadsorbent concentrations, contact time and pH in the adsorption process. Therefore, the XGB proved to be capable to address the predicted adsorption capacity of nanozeolite from a relatively small dataset, being characterized as a starting point before experimental procedures and scale-up of wastewater treatment concerned with organic dye removal.