caa a22 4500 606159398 CHVBK 20210128100624.0 cr unu---uuuuu 210128e20151001xx s 000 0 eng 10.1007/s00521-015-1830-2 doi (NATIONALLICENCE)springer-10.1007/s00521-015-1830-2 Pektas Ali Department of Civil Engineering, Bahcesehir University, 34353, Besiktas, Istanbul, Turkey aut Computational modeling with sensitivity analysis: case study velocity distribution of natural rivers [Elektronische Daten] [Ali Pektas] Determining the velocity profile of an open channel is essential in many hydraulic workspaces such as channel improvement studies, sediment modeling, and energy and turbidity calculations. Since the field observations are labor intensive and time-consuming, many empirical equations have been used for many years. Additionally, many data-based modeling studies have been conducted for both natural rivers and experimental channels. There are two objectives of this study. The first one consists of developing accurate models and criticizing the model performances based on the observational velocity dataset. Hence, classification and regression tree (C&RT), artificial neural network (ANN), and multilinear stepwise regression models are used with different input sets and the models are compared. The second objective is to gain a brief insight about the relationships of the velocity distribution model parameters and determining the significant variables for usage of further modeling studies by considering the co-linearity effects. The relative importance of input variables is investigated on settled models by using sensitivity analysis. The results of the sensitivity analysis indicated that for low-slope natural river studies, instead of using superfluous variables, using only four parameters (U sh , z/H, y/T and z/Y) is adequate to obtain accurate models. The predictive performances of C&RT model and the ANN model were found to be very close to each other, while the multilinear models appeared insufficient. The four variable input set is found superior to other input sets, and the variable water surface velocity is found the most significant parameter across all models. The Natural Computing Applications Forum, 2015 Artificial neural network nationallicence C&RT model nationallicence Multilinear regression models nationallicence Sensitivity analysis nationallicence Natural streams nationallicence Velocity distribution nationallicence Neural Computing and Applications Springer London 26/7(2015-10-01), 1653-1667 0941-0643 26:7<1653 2015 26 521 https://doi.org/10.1007/s00521-015-1830-2 text/html Onlinezugriff via DOI BK010053 XK010053 XK010000 Metadata rights reserved Springer special CC-BY-NC licence nationallicence 1 research-article jats NATIONALLICENCE NATIONALLICENCE NL-springer NATIONALLICENCE 856 40 https://doi.org/10.1007/s00521-015-1830-2 text/html Onlinezugriff via DOI NATIONALLICENCE 100 1- Pektas Ali Department of Civil Engineering, Bahcesehir University, 34353, Besiktas, Istanbul, Turkey aut NATIONALLICENCE 773 0- Neural Computing and Applications Springer London 26/7(2015-10-01), 1653-1667 0941-0643 26:7<1653 2015 26 521