Integration of SVR and GPR Algorithms with Wavelet in Modeling Monthly Drought Forecasting
Subject Areas : Business Administration and EntrepreneurshipJahanbakhsh Mohammadi, 1 , Alireza Vafaeinejad 2 , Saeed Behzadi 3 , Hossein Aghamohammadi 4 , Amirhooman Hemmasi 5
1 - Ph.D. Student, Department of Remote Sensing and GIS, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran.
2 - Associate Professor, Faculty of Civil, Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran.
3 - Assistant Professor, Faculty of Civil Engineering Shahid Rajaee Teacher Training University, Tehran, Iran.
4 - Assistant Professor, Department of Remote Sensing and GIS, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran
5 - Professor, Natural Resources Engineering, Faculty of Natural Resources and Environment, Tehran science and Research Branch, Islamic Azad University, Tehran, Iran.
Keywords: SPI, Drought, SVR, GPR, Wavelet,
Abstract :
Abstract Introduction: Drought is one of the natural hazards that have random and nonlinear behavior due to its various climatic parameters. SPI index is the most common index extracted from rainfall that has been used in modeling drought by various researchers. Methods: The use of computational intelligence methods to model drought in recent years has been much considered by researchers in the field of water resources. In this research, SVR and GPR algorithms individually and also the combination of these algorithms with wavelet algorithms have been modeled and predicted by SPI index, and the purpose was to evaluate the improvement of computational intelligence algorithms in combination with wavelet. In this research, the time series data of 10 synoptic stations in Iran in the period 1961 to 2017 have been used on a monthly basis for modeling the drought as the input of the studied algorithms. Findings: The results of this study showed that the use of the wavelet method in combination with SVR and GPR computational intelligence algorithms improved the results in all time scales. Also, the modeling improvement is due to the use of wavelet in combination with the SVR model with an average RMSE difference of -0.1540 and R2 difference of 0.1491 and the GPR model with an average RMSE difference of -0.1554 and R2 difference of 0.1530 Compared to the single SVR and GPR models showed that the GPR model in general (all time scales and all stations) had a better improvement in the hybrid model than the single model.