Prediction of water level in Kien Giang river using regression-based models

Abstract

     A reliable model to predict the water levels in a river is crucial for better planning to mitigate any risk associated with flooding. Data-driven models using machine learning (ML) techniques have become an attractive and effective approach to model and analyze river stage dynamics. In this study, three
regression-based models, including Linear Regression (LR), Random Forest Regression (RFR) and Light Gradient Boosting Machine Regression (LGBMR) were developed and compared to predict the daily water levels in Kien Giang river based on collected data from 1977 to 2020. Four evaluation criteria, i.e., R2 , NSE, MAE, and RMSE, were employed to examine the reliability of the proposed models. The results show the high accuracy of the proposed models in predicting water levels, especially the LR model. The LR model outperforms the RFR and LGBMR models with the values of R2 , NSE, MAE and RMSE are 0.959, 0.958, 6.67 cm and 12.2 cm respectively.