An integrated machine learning, noise suppression, and population-based algorithm to improve total dissolved solids prediction

Monitoring the water contaminants is of utmost importance in resource management. Prediction total dissolved solid (TDS) particularly essential for quality management and planning areas exposed to a mixture pollutants. TDS primarily includes inorganic minerals organic matters, various salts increasing concentration causes esthetic problems. The reflection pollutant burden aquatic system can remarkably determined by magnitudes. This study focuses on prediction several biochemical parameters such as Na, Ca, HCO3, Mg river system. To overcome nonstationarity, randomness, nonlinearity data, multi-step supervised machine learning evolutionary algorithm (MSMLEA) proposed improve model's performance at two gaging stations, namely Rig-Cheshmeh Soleyman-Tangeh, Tajan River, Iran. In addition, hybrid model that recruits intrinsic time-scale decomposition (ITD) frequency resolution input data well multivariate adaptive regression spline (MARS) were adopted. A novel metaheuristic optimization algorithm, crow search (CSA), was also implemented compute optimal parameter values MARS model. validate model, standalone MARS, empirical mode (EMD)-based models, ITD-MARS MARS-CSA considered benchmark models. Results suggest ITD-MARS-CSA outperforms other