Improving generalization of artificial neural networks in rainfall–runoff modelling

The Artificial Neural Networks are a general-purpose techniques that can be used for non-linear data-driven rainfall-runoff modelling. The key issue to construct a good model by means of Ar-tificial Neural Networks is to understand their structural features and the problems related to their construction. Indeed, the quantity and quality of data, the type of noise and the mathematical properties of the algorithm for estimating the usual large number of parameters (weights) are crucial for the generalization performances of the Artificial Neural Networks. However, it is well known that Artificial Neural Networks may suffer of poor generalization properties due to high number of parameters and non-Gaussian data noise. Therefore, in the first part of the paper, the Artificial Neural Networks features and troubles are discussed. Afterwards eight Avoiding Overfitting Techniques have been presented, considering that Avoiding Overfitting Techniques are methods for improving the Artificial Neural Network’s generalization. For this reason, they have been tested on two cases study, rainfall-runoff data from two drainage basins in the South of Italy, in order to get insight about their properties and to investigate if there is one that absolutely work the best.