Modelling of total dissolved solids in water supply systems using regression and supervised machine learning approaches

Abstract Monitoring of water quality through accurate predictions provides adequate information about management. In the present study, three different modelling approaches: Gaussian process regression (GPR), backpropagation neural network (BPNN) and principal component (PCR) models were used to predict total dissolved solids (TDS) as indicator for The performance each model was evaluated based on sets inputs from groundwater (GW), surface (SW) drinking (DW). GPR, BPNN PCR in this study gave an prediction observed data GW, SW DW, with R 2 consistently greater than 0.850. GPR a better TDS concentration, average , MAE RMSE 0.987, 4.090 7.910, respectively. For BPNN, 0.913, 9.720 19.137, respectively, achieved, while 0.888, 11.327 25.032, assessed using efficiency indicators such Nash Sutcliffe coefficient ( E NS ) index agreement (d). models, (0.967, 0.915, 0.874) d (0.992, 0.977, 0.965). It is understood that advanced machine learning approaches (e.g. BPNN) are appropriate indices would be useful future management parameters various supply systems mining communities where artificial intelligence technology yet fully explored.