Predictive Modeling Approach for Surface Water Quality: Development and Comparison of Machine Learning Models

Water pollution is an increasing global issue that societies are facing and threating human health, ecosystem functions agriculture production. The distinguished features of artificial intelligence (AI) based modeling can deliver a deep insight pertaining to rising water quality concerns. current study investigates the predictive performance gene expression programming (GEP), neural network (ANN) linear regression model (LRM) for monthly total dissolved solids (TDS) specific conductivity (EC) in upper Indus River at two outlet stations. In total, 30 years historical data, comprising 360 TDS EC records, were used models training testing. Based on significant correlation, correlated with seven input parameters. Results evaluated using various measure indicators, error assessment external criteria. simulated outcome indicated strong association actual data where correlation coefficient above 0.9 was observed both EC. Both GEP ANN remained reliable techniques predicting formulated mathematical equations depict its novelty as compared LRM. results sensitivity analysis trend variables affecting HCO3− (22.33%) > Cl− (21.66%) Mg2+ (16.98%) Na+ (14.55%) Ca2+ (12.92%) SO42− (11.55%) pH (0%), while, case EC, it followed (42.36%) SO42−(25.63%) (13.59%) (12.8%) (5.01%) (0.61%) (0%). parametric revealed have incorporated effect all parameters process. criteria confirmed generalized robustness proposed approaches. Conclusively, outcomes this demonstrated formulation AI cost effective helpful river assessment, management policy making.