<![CDATA[Predicting rainfall accurately is essential for managing water resources and preventing hydrometeorological disasters. The unpredictability of daily rainfall patterns necessitates accurate and effective prediction methods. This study proposes a residual hybrid approach to forecast rainfall using historical rainfall data alone. To record temporal information, a Long Short-Term Memory (LSTM) model is used. patterns in the time series data and generate initial predictions. The residual, defined as the difference between actual values and LSTM predictions, is then used as the intend to use a Random Forest (RF) model for training, which learns the non-linear patterns not effectively captured by the LSTM. Although the dataset includes various meteorological variables such as temperature and humidity, this study uses only rainfall as the main input. The data is split into training and testing sets with an 80:20 ratio. Model performance is evaluated using MAE, MSE, and RMSE, with RMSE as the primary evaluation metric. Experimental results show that the LSTM-RF hybrid model consistently delivers greater accuracy of predictions in comparison to single-model approaches, demonstrating strong potential in improving the reliability of rainfall forecasting based solely on historical data.]]>
Copyrights © 2026
