Abstract
Exceptional hydrological events represent one of the most important natural risks which are responsible for the loss of human lives and material goods. During recent decades, many automated or computerized approaches have been implemented to model this process. However, the complexity of hydrological regimes requires the use of specific tools for dynamical and non-linear systems. In order to model the rainfall–runoff transformation, we propose the employment of an adaptive neural network-based fuzzy inference system to predict the flow at the outlet of Algerian coastal basins. The neural network-based fuzzy inference system can be considered as an unlooped neural network for which each layer is a component of a neuro-fuzzy system. The obtained results show that the performances of neuro-fuzzy models exceed those of neural network models and classical multiple linear regression models.
Keywords: