<![CDATA[Weather conditions significantly affect many aspects of modern life, including transportation, tourism, agriculture, and disaster risk management, particularly in relation to rainfall. Consequently, reliable meteorological information is essential for supporting daily decision-making, making rainfall prediction increasingly important. This study develops a daily rainfall prediction model using gradient boosting based on daily meteorological data obtained from the Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG). The dataset includes date, minimum, maximum, and average temperatures, relative humidity, sunshine duration, maximum and average wind speeds, and wind direction, with daily rainfall as the target variable. Four chronological train-test split scenarios were evaluated. The first scenario produced an RMSE of 13.97, an MAE of 7.96, and an R^2 value of 0.14. The second scenario yielded an RMSE of 12.81, an MAE of 8.72, and an R^2 value of 0.17. The third scenario achieved an RMSE of 12.21, an MAE of 7.70, and an R^2 value of 0.20, whereas the fourth scenario obtained an RMSE of 10.31, an MAE of 7.11, and an R^2 value of -0.27. Considering both prediction error and generalization capability, the third scenario was selected as the best-performing model. The main contribution of this study lies in demonstrating the effectiveness of hyperparameter optimization in improving the stability of rainfall prediction under complex tropical climatic conditions. Practically, the proposed model may support BMKG and regional policymakers in Malang Regency in hydrometeorological disaster mitigation and agricultural planning.]]>
