Abstract: Tidal variation has a significant influence on the living pattern and construction development along the coastal region, and thus the prediction of tidal water levels varied with time may become very useful in the design of an engineering project. In this study, a fully year time history tidal recorded data sets at Kaohsiung harbor, the largest harbor in Taiwan were taken as the training basis. The back-propagation neural network and genetic algorithm, with the aid of harmonic tidal equations, were employed to check the reliability of short term wave predictions, as well as to provide a suitable guidance for long term tidal water level forecasting. The result of comparison showed that the approach of neural network has a better performance than that of genetic algorithm, as the former prediction yielded a higher coefficient of correlation and a lower root mean square error in the studied case problem. The predictions including monthly and seasonally wave variations might provide a good reference for relatively engineering design in the investigation area.
Tienfuan Kerh , S.S. Liang and W.G. Chung , 2004. Comparison of Long Term Tidal Water Level Forecasting Using Neural Network and Genetic Algorithm . Asian Journal of Information Technology, 3: 416-427.