Wave height prediction with single input parameter by using regression methods

Abstract

Regression methods can be used for the prediction of parameters under the influence of environmental factors. An effective wave height prediction is important for calculating the wave potential. In the literature, the wave height prediction is generally performed by using the input parameters obtained with different physical effects, with different sensors such as water temperature, daily temperature, daily humidity, wind speed, etc. In this study, the prediction of offshore wave height has been proposed to achieve with a single parameter which is the flow velocity and contains the same physical effects of the wave unlike literature. The effect of different values on different depth of flow velocity has also been investigated by using Relief algorithm. Linear, Decision Tree, Support Vector Machine, Ensemble, and Gaussian Regression models have been studied by using different values of the specified parameters of them for two stations of Mediterranean Sea in Turkey. Various evaluation criteria (MSE, RMSE, MAE, and R_square) have also been utilized to validate the performance of the wave height prediction. The best prediction performances for B-1 and B-2 buoys are obtained with R_square values as 0.866 and 0.954, respectively. These results prove the achievement of the proposed.