Development of SVR-based model and comparative analysis with MLR and ANN models for predicting the sorption capacity of Cr(VI)

Abstract

Abstract Environmental pollution due to heavy metals has become a global concern. Among all existing processes such as chemical precipitation, electro-dialysis etc., biosorption has been recognized as an efficient treatment for the wastewater containing heavy metals like Cr(VI). In this study, the soft computing technique support vector regression (SVR), has been used for predicting the sorption capacity of Cr(VI) with the independent parameters including contact time, initial sorbate concentration, pH of the medium and temperature using agricultural waste ‘maize bran’ as a low cost biosorbent. The developed SVR-based model has been compared with multiple linear regression (MLR) and artificial neural network (ANN) in terms of statistical evaluation parameters. The correlation coefficient (R) for the SVR, ANN and MLR model are 0.9986, 0.9331, 0.8955 while the average absolute relative error (AARE) are obtained as 1.30%, 9.52% and 13.16% respectively. The SVR model is found to be superior than the MLR and ANN models for predicting the sorption capacity of Cr(VI). Furthermore, the effects of the input parameters on the sorption capacity of Cr(VI) employing the MLR, ANN and SVR-based models have been simulated and the obtained results revealed that the SVR-based model is the most accurate, precise, and highly generalized.