دانشگاه ارومیه

 آمار

تعداد نشریات14
تعداد شماره‌ها174
تعداد مقالات1,698
تعداد مشاهده مقاله2,880,503
تعداد دریافت فایل اصل مقاله2,347,846
Amirashayeri, Amin, Behmanesh, Javad, Rezaverdinejad, Vahid. (1404). A Novel Approach for Streamflow Modeling Using Hybrid Machine Learning Techniques. پژوهش های میوه کاری, (), -. doi: 10.30466/jwec.2025.56476.1007
Amin Amirashayeri; Javad Behmanesh; Vahid Rezaverdinejad. "A Novel Approach for Streamflow Modeling Using Hybrid Machine Learning Techniques". پژوهش های میوه کاری, , , 1404, -. doi: 10.30466/jwec.2025.56476.1007
Amirashayeri, Amin, Behmanesh, Javad, Rezaverdinejad, Vahid. (1404). 'A Novel Approach for Streamflow Modeling Using Hybrid Machine Learning Techniques', پژوهش های میوه کاری, (), pp. -. doi: 10.30466/jwec.2025.56476.1007
Amirashayeri, Amin, Behmanesh, Javad, Rezaverdinejad, Vahid. A Novel Approach for Streamflow Modeling Using Hybrid Machine Learning Techniques. پژوهش های میوه کاری, 1404; (): -. doi: 10.30466/jwec.2025.56476.1007

A Novel Approach for Streamflow Modeling Using Hybrid Machine Learning Techniques

Water and Environmental Challenges
مقالات آماده انتشار، پذیرفته شده، انتشار آنلاین از تاریخ 22 مهر 1404
نوع مقاله: Research Article
شناسه دیجیتال (DOI): 10.30466/jwec.2025.56476.1007
نویسندگان
Amin Amirashayeri؛ Javad Behmanesh* ؛ Vahid Rezaverdinejad
Water Engineering Dept., Urmia University
چکیده
This study aimed to predict the daily flow of Zarrineh Rud River using time-lagged inputs and the Artificial Neural Network (ANN) models with a combination of the metaheuristic Firefly Algorithm (FFA) and Reptile Search Algorithm (RSA). The results showed that combining ANN with metaheuristic algorithms consistently improved the prediction accuracy, so that in ANN–RSA1 model at the Sari-Qamish station, the statistics indices were obtained as R² = 0.95 and RMSE = 15.77 m³/s and for ANN–RSA2 model at the Nezam-Abad station, the same indices were calculated as R² = 0.98 and RMSE = 13.21 m³/s both in the test stage. Overall, ANN–RSA delivered the best predictive performance. Increasing time lags improved forecasts at the Nezam-Abad station; however, at the Sari-Qamish station, the input structure to the model was optimized with one time lag, suggesting site-specific lag requirements and potential redundancy when excessive lags are used. The proposed ANN–RSA framework demonstrates high predictive accuracy in arid and semi-arid regions. The findings recommend the application of ANN–RSA for streamflow forecasting and water resources planning, while emphasizing the careful selection of time-lagged inputs to balance complexity and generalization. Future research should evaluate the inclusion of meteorological variables and explore the transferability of the ANN–RSA framework across neighboring basins to strengthen generalizability and support operational decision-making under climate variability, as well as to provide stakeholder-oriented forecasting products for water managers.
کلیدواژه‌ها
Artificial neural networks؛ Reptile search algorithm؛ Time series؛ Urmia Lake
آمار
تعداد مشاهده مقاله: 473