Rainfall Forecasting Using Optimized ML–DL Ensembles
Keywords:
Rainfall Forecasting, Feature Selection, Ensemble Learning, Machine Learning, Deep Learning, Bayesian optimization, LSTM, GRUAbstract
Accurate rainfall prediction is crucial for improving agricultural productivity, reducing flood hazards, and ensuring efficient water resource planning (Ridwan et al., 2021; Wani et al., 2024). Although machine learning and deep learning methods have been extensively employed in rainfall forecasting, model performance is strongly affected by the selection of informative meteorological features, appropriate hyperparameter tuning, and the capability to learn temporal dependencies in climatic data (Aderyani et al., 2022; Chen, 2022). This study conducts a comparative assessment of optimized machine learning and deep learning models for short- and medium-range rainfall prediction. Feature importance is enhanced through mutual information ranking followed by correlation-based feature filtering to remove redundant variables (Johny et al., 2022; Ponnoprat & Phienthrakul, 2021). Random Forest and Support Vector Regression models are optimized using Bayesian hyperparameter search strategies, while Long Short-Term Memory and Gated Recurrent Unit networks and temporal variability inherent in atmospheric systems. Traditional numerical weather prediction models rely on complex are applied to capture sequential rainfall patterns (Li et al., 2023; Xu et al., 2024). Furthermore, a weighted ensemble strategy is introduced to integrate predictions from machine learning and deep learning models, improving forecasting robustness and stability (Anuradha et al., 2024; Wani et al., 2024). Experimental analysis on multi-year meteorological station datasets shows that the ensemble model consistently achieves superior performance compared to individual models across multiple evaluation metrics, particularly during periods of high rainfall variability (Patro et al., 2025; Meteorological Applications, 2025).
References
W. M. Ridwan, A. A. Saputra, and A. M. Idroes, “Rainfall forecasting using
T. Anuradha et al., “Hybrid statistical and machine learning framework for rainfall forecasting,” Expert Systems with Applications, 2024.
F. Alqahtani et al., “LSTM-based deep regression for monthly rainfall forecasting,” Journal of Water and Climate Change, 2024.
M. Dehvari et al., “Rainfall event forecasting using GNSS-derived zenith wet delay and XGBoost,” Advances in Space Research, 2025.
B. S. Patro et al., “Collaborative station learning for regional rainfall prediction,” Atmosphere, 2025.
“Daily precipitation forecasting using long-term station records with ML and DL models,” Meteorological Applications, 2025.
L. Ni, J. Wang, et al., “Hybrid LSTM models for rainfall and streamflow forecasting,” Journal of Hydrology, 2020.
K. Johny, S. Chandran, and A. S. Hameed, “Multivariate EMD–LSTM framework with time-lagged features for rainfall time-series prediction,” Applied Soft Computing, 2022.
F. R. Aderyani, A. H. Nourani, et al., “Comparison of machine learning and deep learning approaches for short-term rainfall prediction,” Journal of Hydrology, 2022.
J. Ponnoprat and P. Phienthrakul, “Season-aware deep learning for daily precipitation forecasting,” Applied Soft Computing, 2021.
D. Li, Y. Zhang, et al., “Hybrid DWT– SVR–Prophet model for rainfall time- series prediction,” Water, vol. 15, no. 10, Art. no. 1935, 2023.
I. Ebtehaj et al., “Comparison of CNN and LSTM for station-based hourly precipitation prediction,” Atmosphere, vol. 15, no. 9, Art. no. 1082, 2024.
H. Xu et al., “Short-term precipitation forecasting using CEEMDAN–GRU hybrid model,” Scientific Reports, 2024.
D. P. Naidu et al., “Monthly rainfall forecasting using SARIMA–BiLSTM hybrid approach,” Journal of Agrometeorology, 2025.
C. Chen, “Monthly rainfall prediction using sliding window LSTM networks,” Tellus A, 2022.
Y. O. Ouma et al., “Trend analysis of rainfall and runoff using LSTM and wavelet neural networks,” Complex & Intelligent Systems, 2022.
H. Wani et al., “Comparative rainfall prediction using ML, DL, and time-series models across altitude zones,” Scientific Reports, 2024.
“Comparative analysis of machine learning models for daily rainfall prediction,” Water Resources Management, 2024.
