Evaluation of artificial intelligence models in river flow modeling, case study: Gamasiab River

Having predicted river flow, we can predict and control natural disasters such as flood and drought in addition to managing utilization of water resources. New models in this domain can help correct management and planning. In this study, three models are evaluated: Gene Expression Planning (GEP), Bayesian Network (BN), and Support Vector Machine (SVM). The data used for this research is precipitation data and daily flow of Gamasiab River in Nahavand during 10 years period (1381-1391). Results indicated that the relative superiority of the gene expression planning model to other models and better performance of SVM model in comparison with BN in daily river flow modeling. In addition, implementing gene expression planning model was faster than other models and could provide results in a short time. The SVM model is also more fitted to estimate the final minimum values. Finally, GEP model with coefficient of determination of 0.9230 and root mean square of 0.5867 in the training phase and coefficient of determination of 0.9025 and root mean square of 0.4936 in the test phase was selected as the superior model.