Forecasting the Water Quality Class in a river basin using an artificial neural network with the softmax activation function

Classification of river water quality needs an efficient method to reduce energy, save time and decrease the risk of errors. This study describes the application of an Artificial Neural Network (ANN) with the softmax activation function to forecast the Water Quality Class (WQC) under the National Water Quality Standard (NWQS) of the Muda River Basin (MRB) (Malaysia). The water quality was classified automatically without Water Quality Index (WQI) calculation. Two different sets of Water Quality Variables (WQVs) were applied as input variables. The modelling discover that the optimal network architecture was the 1:6-1:6-1:1 and used a 60-20-20% splitting plan. ANN1 with the six WQVs was selected to predict the WQC in the MRB. Predictions of the WQC rendered by this model for the training set were very accurate (96.8% correct, Percent Incorrect Prediction (PIP) = 3.2, CEE = 3.44). The approach presented is a very useful and offers a compelling alternative to forecasting of river class, mainly because WQI calculation involves a complex and lengthy calculations. Subsequently, this approach could be applied to water quality classification in other river basins for better water quality management.

Actions (login required)

View Item