Wave height forecasting using cascade correlation neural network

Abstract

Forecasting of wave height is necessary in a large number of ocean coastal activities. Recently, neural networks are used for prediction and approximation of wave heights in sea and ocean due to their great convergence rate. In this paper a cascade correlation neural network is used for prediction of wave heights at given times due to the useful capability of this network for prediction and approximation. Results of different prediction for 500 data points in cascade correlation neural network are compared with those of the M.L.P. (Multi-layer Perceptron) neural network. These results show that cascade correlation network has larger convergence rate compared with M.L.P. network. Also various simulations show that the cascade correlation network has better performance with α=0.005 (Learning-rate), sigmoid activation function for hidden units and linear activation function for output units.