Estimation of Water Quality Parameters With Data‐Driven Model

Abstract

Electrical conductivity and total dissolved solids are considered important parameters in determining quality of drinking and agricultural water because they directly represent total salt concentration in the water. Increases in these parameter values indicate a reduction in water quality. In this study, estimation of the two parameters in the Lighvan Chay River located in Eastern Azerbaijan, Iran, is studied using the k‐nearest neighbors algorithm and support vector regression. Different sets of chemical parameters (i.e., phosphorus, chlorine, calcium, magnesium, sodium, sodium adsorption ratio, sulfate, bicarbonate) were considered as inputs while the total dissolved solids and electrical conductivity were the outputs. Three statistics—coefficient of determination (R2), root mean square error, and mean absolute error—were used to verify accuracy of these models. Comparison of the results showed that both algorithms accurately estimated the total dissolved solids and electrical conductivity, but the support vector regression model is recommended because of better performance.