A comparative study of artificial neural networks and neuro-fuzzy in continuous modeling of the daily and hourly behaviour of runoff

Modeling of rainfall–runoff dynamics is one of the most studied topics in hydrology due to its essential application to water resources management. Recently, artificial intelligence has gained much popularity for calibrating the nonlinear relationships inherent in the rainfall–runoff process. In this study, the advantages of artificial neural networks and neuro-fuzzy system in continuous modeling of the daily and hourly behaviour of runoff were examined. Three different adaptive techniques were constructed and examined namely, Levenberg–Marquardt feed forward neural network, Bayesian regularization feed forward neural network, and neuro-fuzzy. In addition, the effects of data transformation on model performance were also investigated. This was done by examining the performance of the three network architectures and training algorithms using both raw and transformed data. Through inspection of the results it was found that although the model built on transformed data outperforms the model built on raw data, no significant differences were found between the forecast accuracies of the three examined models. A detailed comparison of the overall performance indicated that the neuro-fuzzy model performed better than both the Levenberg–Marquardt-FFNN and the Bayesian regularization-FFNN. In order to enable users to process the data easily, a graphic user interface (GUI) was developed. This program allows users to process the rainfall–runoff data, to train/test the model using various input options and to visualize results.