Comparing Between Support Vector Machine and Gradient Boosted Trees Models for Prediction of Wave Overtopping at Coastal Structures with Composite Slope

Abstract

Background/Objectives: Predicting wave overtopping at coastal structures is a critical task in coastal engineering. The use of machine learning models can help predict wave overtopping with higher accuracy and efficiency. Methods: In this study, the accuracy of support vector machine (SVM) and gradientboosted tree (GBT) approaches for predicting wave overtopping discharge of coastal structures with composite slopes "without a berm" was evaluated and compared. The newly developed EurOtop database was used for this study. Findings: The results showed that the GBT technique provided more accurate predictions than the SVM technique. Notably, the results demonstrated that the GBT model significantly decreased the overall error and accurately estimated the wave-overtopping discharge. Compared to the SVM model, the scatter index calculated using the GBT (0.57) is lower than that of the SVM (o.94). In terms of R-index, the GBT (0.97) was superior to the SVM (0.92) model. Novelty: This study provides an effective approach for predicting wave overtopping discharge in the design and management of coastal structures.