Abstract
Given the increasing recognition of machine learning tools for use in water quality monitoring, enhancing their applicability in full-scale plants require investigation of their capabilities and limitations in key aspects of the water supply chain. This study comprehensively evaluates the performances of three Artificial Neural Network (ANN) training algorithms and three solvers for regression Support Vector Machine (SVM) with different kernel functions in the estimation of the counts of Coliform bacteria from measured records of physcho-chemical water quality parameters. In addition, input data were subjected to different normalization methods to determine their effects on the performances of both ANN and SVM models. The feedforward and the cascade forward algorithms yielded the lowest MSE values among the various ANN model configurations. No distinct disparity was found in the performances of the various solvers of regression SVM in the estimations. For the regression SVM kernel functions, the Radial Basis Function (RBF) and the Gaussian kernel functions resulted in the lowest MSE values. Both the ANN and regression SVM have comparable abilities in predicting the levels of the faecal indicator organisms in the raw water. However, the ANN models were more efficient in estimating intense variations in the levels of the indicator organisms in raw water. DOI: http://dx.doi.org/10.5755/j01.erem.74.1.20083
Highlights
The wide range of applications of machine learning algorithms and computational intelligence as decision support tools has made them indispensable in the water supply industry today
By removing turbidity from the artificial neural network (ANN) models (Fig. 5 A), the mean square error (MSE) values increased from 2.02 CFU/100 mL to 2.47 CFU/100 mL and from 2.48 CFU/100 mL to 2.95 CFU/100 mL, respectively, for coliform bacteria and E. coli predictions
The water quality parameter that showed the least sensitivity in the predictions was the water pH, with MSE increases of approximately 3% (ANN models) and 2% (SVM models), and the changes were similar in both coliform bacteria and E. coli prediction models
Summary
The wide range of applications of machine learning algorithms and computational intelligence as decision support tools has made them indispensable in the water supply industry today. Researchers have applied various data-driven techniques to explain the influences of various physico-chemical water quality parameters on concentrations of faecal indicator organisms (FIOs) and other water quality parameters in raw water These mostly include regression methods (Black et al, 2007; Juntunen et al, 2012) and artificial intelligence methods such as artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS) and support vector machine (SVM) (Singh et al, 2009; Kim et al, 2012; Heddam, 2014; Mohammed et al, 2017). Other studies compared the performance of regression and ANN, ANFIS and SVM methods in the prediction of water quality parameters (Abyaneh, 2014; Chandramouli et al, 2007; Zhang et al, 2015) and reported higher accuracy of these methods relative to conventional regression methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Environmental Research, Engineering and Management
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
