Integrating machine learning and statistical design for sustainable Fenton catalysis: enhancing dye degradation in wastewater treatment

The use of computer modeling in the Fenton process holds considerable promise for enhancing process efficiency and sustainability. Although considerable study has been conducted on the traditional Fenton process, investigations using simulation techniques remain very limited. This study introduced an innovative method that integrates experimental design with machine learning to improve predicted accuracy. A locally weighted kernel partial least squares regression (LW-KPLSR) model was developed using Taguchi's orthogonal array architecture to forecast dye degradation efficiency across diverse operating situations. The evaluation of model performance was conducted using root mean square error (RMSE), mean absolute error (MAE), and the coefficient of determination (R-2). For comparison analysis, supplementary regression models-principal component regression (PCR), locally weighted partial least squares regression (LW-PLSR), partial least squares regression (PLSR), least squares support vector regression (LSSVR), and fuzzy modeling were used. As compared between the RMSE values of LW-KPLSR and LSSVR, LSSVR outperforms 39% to 329% in all case studies. Among all the models, LSSVR demonstrated the highest predictive ability, as indicated by significantly reduced RMSE and MAE values, along with a high R-2 value of 0.9887.