Comparing Nonlinear Time Series Models and Genetic Programming for Daily River Flow Forecasting (Case study: Barandouz-Chai River)

Document Type : Complete scientific research article

Authors

Abstract

Abstract
In this study daily river flow of Barandouz-Chai in the period of 1973 to 2009 has been forecasted using nonlinear time series and Genetic programming models and result have been analyzed with root mean square error and regression coefficient methods. After evaluating model by goodness of fit, nonlinear BL(1,11,1,1) with the minimum AICC selected for modeling daily river flow forecasting. For BL(1,11,1,1) model regression coefficient and RMSE calculated equal to 0.902 and 3.52 (m3/s) respectively. Genetic programming has been used for modeling river flow with consideration to memory of one, two, three and four days before. Results showed until three days memory accuracy of model is acceptable but after that accuracy of model decreased. Regression coefficient and RMSE of Genetic programming in the testing phase calculated equal to 0.928 and 2.863 (m3/s) respectively. As results showed, Genetic programming with 22.9 percent less error was better than Bilinear time series model in Barandpuz-chai river flow forecasting

Keywords

Files

Share

How to cite

Statistics