Forecasting of Wave Height Time Series Using AdaBoost and XGBoost, Case Study in Pangandaran, Indonesia

Wave height forecasting has a vital role in coastal activities. The prediction results help minimize losses on various ocean activities, such as sea transportation and designing coastal structures. The numerical method is usually utilized for wave height forecasting, but it requires high computational resources to solve its complexity. Moreover, the high-resolution grid resolution is also required to produce an accurate prediction. In this paper, we use XGBoost and AdaBoost methods to forecast the wave height accurately. Five years of data is used in this study, 4.5 years of data as training data, and six months of data as testing data. We use the testing data to forecast 7 days, 14 days, 30 days, 45 days, and 60 days ahead. We use wave data at the southern part of the Pangandaran region, Indonesia, as a study case. We compared the forecasting results by using XGBoost and AdaBoost to obtain the highest accuracy. Hyperparameter tuning for both methods is also used in this study to optimize the model's performance. We use the GridSearch algorithm for hyperparameter tuning to find the best parameters value combination for both models. The highest accuracy of forecasting results is 0.064 for RMSE and 0.987 for the coefficient correlation value.