Automated machine learning approach for accurate prediction of adsorptive removal of dye in a coconut fiber packed column system

The paucity and demand of pure drinking water awaken the need of developing effective wastewater treatment strategies. One of the leading contributors of water pollution are dyes which are released into water bodies through industrial effluents. The present study aims to evaluate the potential of coconut fibers (CF) to adsorb crystal violet (CV) dye through column adsorption and develop an accurate machine learning (ML) model for predicting the adsorption performance. The CF adsorbent displayed an extraordinary removal of 100% for 13 L of dye solution under investigate operating conditions (flow rate = 4 to 10 mL/min, bed height = 4.5 cm, inlet dye concentration = 50 mg/L, pH = 8). Furthermore, artificial intelligence (AI) and auto machine learning (ML) techniques have been applied on the data set, for prediction of adsorption percentage. The unmapped data and the relationship between the input and output variables have been studied by various visualization techniques like histograms, box plots, pair plot, Pearson’s correlation, and mutual information gain. Asynchronous successive halving algorithm (ASHA), an auto ML technique for hyperparameter optimization has been employed for selecting the best suitable model. Among the evaluated models, support vector machine (SVM) was found to best performing model with high R 2 value of 1. The reliability and generalizability of the selected model were further examined through residual-error analysis and cross-validation, and the limitations related to the dataset size and the limited range of operating conditions employed were discussed in detail. Overall, the research work highlights the utilization of low-cost biosorbent for large scale dye removal and demonstrates the effective integration of auto ML approach as a reliable tool for accurate prediction of adsorption capacity of adsorbents in the domain of environmental remediation.