Accuracy Comparison of Random Forest Method Against ANN MLP for Forecasting Significant Wave Height for Alghero, Italy

Abstract

Around 1 year of measured data for Alghero, Italy was used to train and test two machine learning algorithms namely the Random Forests method and the Multi-Layer Perceptron (MLP) with the Random Forest being the best. The input predictors were the wind velocity, wave peak period and the significant wave height (SWH). These were normalized before being fed to the model. The output response was significant wave height. It is difficult to select wave energy extraction sites therefore a better understanding of significant wave height is needed through accurate prediction. This work explores the possibilities of exploring and comparing machine learning algorithms which could forecast SWH accurately since it is the most important parameter used in studying the energy output of wave energy converters.