Assessment of water surface profile in nonprismatic compound channels using machine learning techniques

Accurate prediction of water surface profile in an open channel is the key to solving numerous critical engineering problems. The goal of the current research is to predict the water surface profile of a compound channel with converging floodplains using machine learning approaches, including Gene Expression Programming (GEP), Artificial Neural Networks (ANN), and Support Vector Machines (SVM), in terms of both geometric and flow variables, as past studies were more focused on geometric variables. A novel equation was also proposed using gene expression programming to predict the water surface profile. In order to evaluate the performance and efficacy of these models, statistical indices are used to validate the produced models for the experimental analysis. The findings demonstrate that the suggested ANN model accurately predicted the water surface profile, with coefficient of determination (R2) of 0.999, root mean square error (RMSE) of 0.003, and mean absolute percentage error (MAPE) of 0.107%, respectively, when compared to GEP, SVM, and previously developed methods. The study confirms the application of machine learning approaches in the field of river hydraulics, and forecasting water surface profile of non prismatic compound channels using a proposed novel equation by gene expression programming made this study unique.