PREDICTION OF DAILY TIDAL LEVELS ALONG THE CENTRAL COAST OF EASTERN RED SEA USING ARTIFICIAL NEURAL NETWORKS

Abstract

Accurate tidal prediction is essential for the design and construction of coastal and marine
structures. In this study, an Artificial Neural Network (ANN) approach uses four algorithms (Radial Basis
Function, General Regression, Multilayer Perceptron, and Cascade Correlation) were developed to estimate
the tidal levels along the central coast of eastern Red Sea. A genetic algorithm was used to determine the
adequate ANN structure and the optimal values of the parameters for the different algorithms of the ANNs.
The obtained results confirm that the General Regression Neural Network (GRNN) model outperforms the
other techniques. Moreover, the results verify that the GRNN model provides improvements in root mean
square errors of 117.15%, 122.85%, 121.43%, and 127.15% over the Multilayer Perceptron Neural Network
(MPNN) with three layers, MPNN with four layers, Cascade Correlation Neural Network (CCNN), and Radial
Basis Function Neural Network (RBFNN), respectively for training and 19.26%, 20.50%, 11.8%, and 23.61%
for testing. This investigation further indicates that the GRNN model can be useful as a supervised learningbased tool for predicting tidal levels.