Evaluation the efficiency of radial basis function neural network for prediction of water quality parameters

This manuscript attempts to acquaint with the suitability of Artificial Neural Networks (ANN) especially Radial Basis Function RBF to predict water quality parameters. Search for optimal model parameters within RBF-NN is carried out in two steps, each of which can be made to be more efficient and much faster than in MLP. This study focused on electrical conductivity, total dissolved solids and turbidity as the main water quality parameters for rivers. The employed model proved its capabilities to mimic the inter-relationship between such water quality parameters. Normalization and partitioning for the row data have been carried out to accelerate the training process and to achieve pre-defined Sum Square Error SSE equal to 10(-4). The results show that the proposed RBF-NN after normalization and partitioning outperformed the linear regression model and achieve Mean Absolute Prediction Error MAPE equal to 8.3%.