Performance Analysis of Neural Network Architecture in Developing Real-Time Malaysian River Water Quality Model

Abstract According to the world health organization, 485,000 people died each year due water-related diseases which are mainly contributed by poor river water quality. As a result, quality monitoring stations have been deployed across world. Unfortunately, complex nature of off-site parameters, index (WQI) cannot be assessed in real-time. This has led significant push for scientific community develop an accurate and robust prediction model. The dynamic nonlinear parameters major challenges traditional machine learning algorithms such as multi linear regression capture. In this study, model was developed using feedforward artificial neural network (FANN) utilizing only on-site parameters. performance different activation functions hidden neurons thoroughly analysed includes rectified unit (ReLU), scaled exponential (SELU), (ELU). Additionally, various initialization optimization were also evaluated maximum efficiency. results shows that FANN-ELU coupled with Glorot technique AdaGrad optimizer outperformed other combinations R 2 value 0.88 mean squared error (MSE) 22.74.