Prediction of the five-day biochemical oxygen demand and chemical oxygen demand in natural streams using machine learning methods

Abstract

Rivers, as the most prominent component of water resources, have a key role to play in increasing the life expectancy of living creatures. The essential characteristics of water pollutants can be described by water quality indices (WQIs). Hence, a ferocious demand for obtaining an accurate prediction of WQIs is of high importance for perception of pollutant patterns in natural streams. Field studies conducted on different rivers indicated that there is no general relationship to yield water quality parameters with a permissible level of accuracy. Over the past decades, several artificial intelligence (AI) models have been employed to predict more precise estimation of WQIs rather than conventional models. In this way, through the current study, multivariate adaptive regression spline (MARS) and least square-support vector machine (LS-SVM), as machine learning methods, were used to predict indices of the five-day biochemical oxygen demand (BOD5) and chemical oxygen demand (COD). To improve the proposed approaches, 200 series of field data, collected from Karoun River southwest of Iran, pertain to the nine independent input parameters, namely electrical conductivity (EC), sodium (Na+), calcium (Ca 2+), magnesium (Mg 2+), orthophosphate ([Formula: see text]), nitrite ([Formula: see text]), nitrate nitrogen ([Formula: see text]), turbidity, and pH. The performances of the LS-SVM and MARS techniques were quantified in both training and testing stages by means of several statistical parameters. Furthermore, the results of the proposed AI models were compared with those obtained using artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and multiple regression equations. Results of the present research work indicated that the proposed artificial intelligence techniques, as machine learning classifiers, were found to be efficient in order to predict water quality parameters.