Olufunminiyi Abiri1,*, Benjamin Oyediran Oyelami1, Chindo Istifanus Yarkasuwa2, Buba Mamman Wufem3
1National Mathematical Centre, Abuja 900001, Nigeria.
2Department of Chemistry, Abubakar Tafawa Balewa University, PMB 0248, Bauchi, Nigeria.
3Department of Chemistry, Plateau State University, PMB 2012, Bokkos, Nigeria.
*Corresponding author: Olufunminiyi Abiri
Abstract
In this paper, a Multilayer Feedforward Neural Network model with Bayesian regularization (Levenberg-Marquardt) is developed as a means of predicting water points quality parameters in Nigeria. The use of Bayesian regularized technique reduces the potential of overfitting and overtraining, thereby improving the prediction quality of the model. The MFNN model, uses six input variables identified as key high-risk parameters influencing Total Dissolved Solids (TDS) and these are: Total Coliform (TTC), Cadmium (Cd), Nitrate (NO3-), Fluoride (F), Arsenic (As), and Lead (Pb) concentrations. The output variable TDS was selected as the predicted variable, serving as a general indicator of inorganic water pollution. The complete dataset (n=225) was randomly partitioned into training (70%, n=158), validation (15%, n=34), and testing (15%, n=33) sets. All input variables were normalized to [0,1] range using min-max scaling. The training set was used for weight optimization, the validation set for monitoring overfitting and determining early stopping, and the test set for the final, unbiased evaluation of model performance. All input variables were normalized prior to training. MFNN training with the dataset, shows mean square error (MSE) decreases as a function of the number of epochs. At convergence, the MSE error is 0.000401, and a high correlation (R=0.97) between predicted and observed TDS, demonstrating its potential for predicting water quality parameters above regulatory limits. The model was able to predict the TDS values to be 985.00 which is above the Nigeria Standard for Drinking Water Quality Maximum Permissive Level of 500.
References
[1] Oyelami O, Wufem B. Models for Computing Emission of Carbon Dioxide from Liquid Fuel in Nigeria. American Journal of Mathematical and Computer Modelling. 2017;2(1):29-38. doi:10.11648/j.ajmcm.20170201.15.
[2] Oyelami O. Models for Computing Effect of Pollutants on the Lower Respiratory Tract. American Journal of Modelling and Optimization. 2016;4(2):40-50. doi:10.12691/ajmo-4-2-2.
[3] Zannetti P, editor. Air pollution modeling: theories, computational methods and available software. Springer Science & Business Media; 2013.
[4] Khan I, Raja MAZ, Shoaib M, Kumam P, Alrabaiah H, Shah Z, et al. Design of neural network with Levenberg-Marquardt and Bayesian regularization backpropagation for solving pantograph delay differential equations. IEEE Access. 2020;8:137918-137933.
[5] Waldo J. A comparative study of back propagation and its alternatives on multilayer perceptrons. arXiv preprint arXiv:2206.06098. 2022.
[6] Abiri O, Twala B. Modelling the flow stress of alloy 316L using a multi-layered feedforward neural network with Bayesian regularization. In: 2017 2nd International Conference on Knowledge Engineering and Applications (ICKEA). 2017. p. 80-84. IEEE.
[7] Babayemi JO, Ogundiran MB, Osibanjo O. Overview of environmental hazards and health effects of pollution in developing countries: a case study of Nigeria. Environ Qual Manag. 2016;26(1):51-64. doi:10.1002/tqem.21477.
[8] Nnaemeka AN. Environmental pollution and associated health hazards to host communities (Case study: Niger delta region of Nigeria). Central Asian Journal of Environmental Science and Technology Innovation. 2020;1(1):30-42.
[9] Vigil KM. Clean water: an introduction to water quality and water pollution control. 2nd ed. Corvallis, Oregon: Oregon State University Press; 2003.
[10] Jospin LV, Laga H, Boussaid F, Buntine W, Bennamoun M. Hands-on Bayesian neural networks—a tutorial for deep learning users. IEEE Comput Intell Mag. 2022;17(2):29-48.
[11] Sterratt D, Graham B, Gillies A, Einevoll G, Willshaw D. Principles of computational modelling in neuroscience. 2nd ed. Cambridge University Press; 2023.
[12] Duma IS, Twala B, Marwala T. Predictive modeling for default risk using a multilayered feedforward neural network with bayesian regularization. In: The 2013 International Joint Conference on Neural Networks (IJCNN); 2013 Aug 4-9; Dallas, TX, USA. IEEE; 2013. p. 1-10.
[13] Sariev E, Germano G. Bayesian regularized artificial neural networks for the estimation of the probability of default. Quant Finance. 2020;20(2):311-28.
[14] Lampinen J, Vehtari A. Bayesian approach for neural networks—review and case studies. Neural Netw. 2001;14(3):257-74.
[15] Rumelhart DE, Durbin R, Golden R, Chauvin Y. Backpropagation: the basic theory. In: Backpropagation: theory, architectures, and applications. Psychology Press; 2013. p. 1-34.
[16] Beale MH, Hagan MT, Demuth HB. Neural Network Toolbox™ getting started guide. Natick, MA: The MathWorks, Inc.; 2016.
[17] Farooq MU, Zafar AM, Raheem W, Jalees MI, Aly Hassan A. Assessment of algorithm performance on predicting total dissolved solids using artificial neural network and multiple linear regression for the groundwater data. Water. 2022;14(13):2002. doi:10.3390/w14132002.