An optimized XGBoost-based machine learning method for predicting wave run-up on a sloping beach.

Area of Science:

• Coastal engineering
• Machine learning applications
• Hydrodynamics

Background:

• Accurate wave run-up prediction is crucial for mitigating coastal inundation and erosion from various oceanic events.
• Traditional methods like physical experiments and numerical modeling have limitations in efficiency and applicability.
• Machine learning offers a robust approach for complex data analysis in wave run-up modeling.

Purpose of the Study:

• To introduce an extreme gradient boosting (XGBoost)-based machine learning method for predicting wave run-up on sloping beaches.
• To optimize the XGBoost model through hyperparameter tuning using a grid search approach.
• To compare the performance of the optimized XGBoost model against other machine learning techniques and empirical formulas.

Main Methods:

• Utilized over 400 laboratory observations of wave run-up as training data.
• Developed an XGBoost model and optimized its hyperparameters via grid search.
• Compared XGBoost performance with Multiple Linear Regression (MLR), Support Vector Regression (SVR), and Random Forest (RF).

Main Results:

• The optimized XGBoost model achieved a high prediction accuracy with a correlation coefficient (R) of 0.98675.
• Achieved a Mean Absolute Percentage Error (MAPE) of 6.635% and a Root Mean Squared Error (RMSE) of 0.03902.
• Demonstrated superior performance compared to MLR, SVR, and RF models.

Conclusions:

• The optimized XGBoost method is a feasible and highly accurate alternative to empirical formulas and classical numerical models for wave run-up prediction.
• The XGBoost model shows broader applicability across various beach slopes and incident wave amplitudes than traditional empirical formulas.
• XGBoost offers a computationally efficient solution compared to numerical models for wave run-up assessment.