Normalisation methods on neural networks for predicting pavement layer moduli

An artificial neural networks-based model has recently been proposed for predicting pavement layer moduli and it is shown that this model has the potential to achieve better performance in terms of computing cost and prediction accuracy than the linear elastic layered theory commonly used in pavement layer backcalculation. Even though the backpropagation neural network could be a successful model for predicting pavement layer moduli, there are some major issues that need to be considered before using the neural network models, such as the network topology, learning parameter, and normalisation methods for the input vectors. In this research, five normalisation methods for input vectors were studied and the experimental results showed that the neural network performance for predicting pavement layer moduli was highly dependent on the normalisation methods. In addition, the best normalisation method for predicting pavement layer moduli was suggested in order to achieve the best prediction performance as well as to reduce the computing cost for training.