Comparison of Artificial Intelligence Algorithms in Daily River Flow Modeling

Subject Areas : Water Resources

Massome  Zeinali ¹ (Graduated Master of Water Resources Engineering, Faculty of Agriculture, Razi University, Kermanshah.)
Sohila  Farzi ² (Graduated Master of Irrigation and Drainage, Department of Water Engineering, Faculty of Agriculture, Razi University, Kermanshah, Iran.)
Mohammad Reza  Golabi ³ (Ph.D. Student, Department of Water Resources Engineering, Faculty of Water Sciences, Shahid Chamran University of Ahvaz, Iran.)
feridon  radmanesh ⁴ (Associate Professor of Water Engineering Faculty of Water Sciences, Shahid Chamran University of Ahvaz, Iran.)

Received: 2017-12-12 Accepted : 2018-01-13 Published : 2017-12-22

Keywords: ANFIS model, Stream modeling, BN model, SVM model,

Abstract :

One of the most important issues in water resources engineering is the prediction of river flow rates as one of the main sources of human water supply, which is important in terms of water resources planning. Using new models in this field can help to manage and plan correctly. In this study, we evaluated 3 models called Neural-Fuzzy Network (ANFIS), Busin Network (BN) and Backup Machine Vector (SVM). The data used for this research is precipitation data and daily flow of Gamasiab Nahavand River during a 10 year period (1381-1391). The results indicated that the neural-fuzzy network model (ANFIS) and backup vector machine (SVM) had almost the same performance in daily river flow modeling and had better performance than the network model. In addition, the speed of implementation of SVM model compared to the rest The models were bigger and were able to deliver results in a short time.