Error-replaced recurrent randomized-based ensemble algorithm for water level prediction

Zongying Liu

doi:10.1016/j.jhydrol.2026.135668

Error-replaced recurrent randomized-based ensemble algorithm for water level prediction

2026-08-01

0

PRE

AI

DOI：10.1016/j.jhydrol.2026.135668

Original

Original request for help

Save

Abstract

En 中文

With the rapid development of inland waterways and coastal management, accurate water level prediction has be come increasingly important. Reliable forecasting of water levels is essential for flood mitigation, water resource management, and the operation of inland waterways and ports. However, challenges remain due to algorithmic limitations and the complex dynamics of natural water systems. This study introduces a novel algorithm, the Error-Replaced Recurrent Random Ensemble Algorithm (ER-REA), designed to enhance both accuracy and stability in water level prediction. The proposed method constructs a randomized ensemble architecture and incorporates an error-replaced recurrent mechanism that dynamically updates the output weights at each forecasting step, thereby mitigating error accumulation in multi-step prediction. The performance of ER-REA is evaluated against several baseline models, including Extreme Learning Machine, Echo State Network, and AdaBoost Extreme Learning Machine, across seven real-world water level datasets. Experimental results demonstrate that ER-REA consis tently achieves superior predictive accuracy and stability across multiple forecasting horizons, reducing average error by 37.06% compared with the best-performing baseline model. Comparisons with conventional machine learning and deep learning models further demonstrate the competitiveness of ER-REA. In summary, the pro posed model provides a reliable tool for water level prediction, contributing to improved flood risk mitigation and sustainable port operations.

Keywords:

Random ensemble structure

Randomization-based model

Error-replaced recurrent algorithm

Water level prediction

Error-replaced recurrent randomized-based ensemble algorithm for water level prediction

Abstract

Journal

Organization

Cited Papers