Prediction of wave overtopping discharges at coastal structures using interpretable machine learning

ABSTRACT

Appropriate estimation and prediction of wave overtopping discharges are very important in terms of economics, port structure stability, and port operation. In recent years, machine learning (ML) techniques, which predict by finding statistical structures from input/output data using computers, have generated interest. However, as the complexity of ML models increases, interpreting their results becomes increasingly difficult. Interpretation of ML results is an important part in developing an efficient structure design strategy for improved wave overtopping discharge estimation. Therefore, in this study, eight linear/nonlinear ML models were applied to the same data, and a pipeline model for selecting an ML model suitable for data characteristics was developed. In addition, the importance of variables related to the prediction of wave overtopping discharges and their correlations were analyzed by interpretable ML. The research results showed that the extreme gradient boosting model had the highest prediction accuracy and significantly reduced the error. Accordingly, a data-based model can be a new alternative for analyzing the complex physical relationships in the field of coastal engineering and used as a starting point toward structure design and development for coastal disaster prevention.

KEYWORDS:

• overtopping
• prediction
• machine learning
• feature engineering
• coastal engineering

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Publicly available datasets were analyzed in this study. The data can be found here: http://www.overtopping-manual.com/eurotop/downloads/

Additional information

Funding

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2022R1C1C2004838). This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. NRF-2021R1A2C4002665).