Predicting Quality of River's Water Based on Algae Composition Using Artificial Neural Network

Artificial neural networks which are inspired by the concept of the biological neurons are commonly used in many applications including in the field of water quality management. The neural network approaches have provided an educated solution to aid in the decision-making process for river system as well as a viable means of the forecasting for water quality parameters. This paper attempts to determine the suitability and the applicability of artificial neural networks for detecting quality of river's water based on algae composition. 21 different types of algae have been used as input data and the river's water was classified into 4 categories, namely clean, polluted, brackish and moderate. Multilayered perceptron network with three different learning algorithms have been studied. The multilayered perceptron trained using Bayesian Regularization algorithm has been proven to produce the best results with high accuracy percentage (93.50%) as compared to the Lavenberg Marquadt (93.00%) and back propagation (63.505%). Further analysis (i.e. more testing data, new architecture of neural network) will be carried out to further improve the system.