Data-driven approaches in predicting scour depths at a vertical seawall on a permeable shingle foreshore

Abstract Despite recent research efforts in advancing machine learning (ML) tools to predict nearshore characteristics at sea defences, less attention has been paid ML algorithms predicting scouring vertical seawalls. In this study, four approaches were investigated, including gradient boosting decision trees (GBDT), random forest (RF), support vector regression (SVR), and ridge (RR). These utilised scour depths the toe of an impermeable seawall front a permeable shingle slope. The developed trained tested (70% for training 30% testing) using datasets collected from laboratory tests performed on seawalls 2D wave flume University Warwick. A novel hyperparameter tuning analysis was each model tailor underlying dataset features while mitigating associated data overfitting risks. Additionally, process demonstrated permutation feature importance reduce redundancy. predictions compared with observed values coefficient determination ( R 2 ) score, root mean square error (RMSE), Pearson correlation -value. Consequently, RF GBDT methods accurately predicted foreshores. This study produced data, information, that could directly or indirectly benefit coastal managers, engineers, local policymakers. benefits included forecasting assessing impact structural integrity defences response threat imposed by extreme events, which are essential sustainable management protections properties behind such structures areas.