Answers are grounded in this paper and cite their sources.
AI models enhance prediction accuracy for flood forecasting in Sarawak River, suggesting new water management strategies.
River water levels are influenced by complex and dynamic mechanisms, posing significant challenges for accurate predictions, especially in flood-prone areas. To address this issue and enhance flood forecasting capabilities, this study employed four AI models—LSTM, BiLSTM, CNN-BiLSTM, and CNN-BiLSTM-XGB—to analyze river water level data of the Sarawak River. While the first three models performed well on training and validation sets, their test set results were overfitted, with R2 values failing to reach 0.9. To improve performance, the CNN-BiLSTM-XGB model was introduced. This hybrid model effectively captures complex data patterns, accelerates convergence, and enhances training stability and efficiency. Experimental results demonstrated that the CNN-BiLSTM-XGB model outperformed others on both validation and test sets, achieving an R2 value of 0.973, a 17.1% improvement over the LSTM model. Extensive simulations confirmed the model’s superior prediction accuracy and robustness. Overall, the CNN-BiLSTM-XGB model proved to be a highly effective tool for short-term river level prediction, offering significant potential for improving flood forecasting and water resource management strategies.
No takes yet. Share an insight, caveat, or question.
H. C. Zhou (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: