|
Machine Learning-Based Streamflow Discharge Forecasting
|
Sayyad Asghari Saraskanroud *   , Amin Abdolalipouradl |
| Professor ، Department of Physical Geography ، Faculty of Social Sciences ، University of Mohaghegh Ardabili ، Ardabil ، Iran |
|
|
Abstract: (19 Views) |
| According to the Gara su is the most significant Permanent rivers in Ardabil and plays a central role in supplying and watering in management agriculture and gardens. This study employed five machine learning models—namely Random Forest (RF), k‑Nearest Neighbors (KNN), Gradient Boosting, Artificial Neural Network (ANN), and Linear Regression (LR) were employed to predict daily river discharge at two hydrometric stations, Samian and Arbab Kandi. In the first step, missing river discharge and climatic data were removed. In addition, for more specific processing, a 7‑day average discharge and a 3‑day average rainfall was considered. Additionally, we divided data into three parts including training 70 % validation 15% and testing 15%. Moreover, results showed that at the Samian Station the Gradient Boosting model achieved the best performance in coefficient of determination (R²) of 0.8617 and a root mean square error (RMSE) of 0.631. In contrast, at the Arbab kandi station, the Random Forest model with the best score an R20.7461 and RMSE about 1.1726, but also the KNN model at both stations recorded low R2. an R2 0.7207 and 0.5340. In addition, RMSE value was about 1.2760 and1.1597. Finally, the KNN was the worst model among machine learning models in this study. |
Article number: 4 |
| Keywords: Ardabil, Machine learning, discharge, prediction |
|
|
|
Type of Study: Research |
Subject:
Special
Received: 2026/07/8 | Revised: 2026/09/18 | Accepted: 2026/09/18 | ePublished ahead of print: 2026/09/18
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
