Predictive modeling of thiosulfate removal from wastewater using the machine learning models XGBoost and SVR

Abstract

ABSTRACT The separation of thiosulfate from wastewater is important to prevent environmental pollution and ensure water quality compliance. In this study, two widely used machine learning (ML) models, namely extreme gradient boosting (XGBoost) and Support Vector Regression (SVR), have been applied to predict concentration of thiosulfate. Experimental data based on elimination of thiosulfate from wastewater using aerial oxidation and photooxidation with and without ultraviolet light under varied experimental conditions was used for ML modeling. Time-series-based input variables included the initial concentration of thiosulfate, oxygen level, air-flow rate, and ultraviolet light intensity. During data preparation for models’ development lags of thiosulfate concentrations at previous time stamps were created. The data was normalized, and grid search cross-validation method was utilized for hyperparameter optimization. XGBoost model gave the Root Mean Squared Error (RMSE) value as 47.729 with (R-squared) R2 value of 0.970 with the test set. Similarly, SVR model demonstrated the RMSE value as 61.904 with R2 value of 0.902 with the test set. The results demonstrate the potential of ML models in the optimization of parameters and enhancement of efficiency for the wastewater treatment.