Fig 1.
Schematic of parameters used in describing wave overtopping discharge at sea defences (adopted from [3]).
See glossary for further description of the parameters.
Fig 2.
Typical structure of a decision tree (DT).
Fig 3.
The structure and workflow of a RF algorithm.
Fig 4.
The structure and workflow of a GBDT algorithm.
Fig 5.
The structure of a SVR algorithm.
Fig 6.
Structure of a feed forward back propagation ANN.
Table 1.
List of the hyperparameters and their corresponding best values.
Table 2.
Range of overtopping parameters used to extract data for vertical breakwaters.
Table 3.
Correlation table of overtopping parameters.
Fig 7.
PCA Decomposition and scores of the dataset.
Fig 8.
Framework of the methodology.
Fig 9.
Comparison of the dimensionless q (predicted) and dimensionless q (measured)
for—a) RF, b) ANN, c) GBDT, and d) SVR.
Table 4.
Evaluation scores of ML algorithms for five key statistical metrices.
Fig 10.
a) Feature importance of the RF and GBDT and b) Permutation Importance analysis of SVR and ANN.
Fig 11.
Variation of dimensionless overtopping rate ‘q’ with Rc/Hm0,t for a) RF, b) ANN, c) GBDT, and d) SVR.
Fig 12.
Variation of discrepancy ratio (DR) with Rc/Hm0,t for a) RF, b) ANN, c) GBDT, and d) SVR.
Fig 13.
Taylor’s diagram of the statistical scores from the prediction analysis of the four ML algorithms.