<![CDATA[This study aims to compare the performance of the Seasonal Autoregressive Integrated Moving Average (SARIMA) and Adaptive Momentum Backpropagation (AMBP) models for predicting monthly rainfall in West Nusa Tenggara Province. The data used consists of monthly rainfall from the NASA POWER dataset for the period from January 1981 to December 2025. The evaluation was conducted using a block-based forecasting approach that divides the data into eight temporal blocks with a 120-month training window and a 12-month forecast horizon. Model performance was evaluated using several evaluation metrics to measure the models’ accuracy and generalization ability. The results show that the SARIMA model has better and more stable predictive performance, with an average RRMSE of 49.18% and (R²) of 68.13%, while the AMBP model yields an average RRMSE of 62.83% and an R² of 42.40%. Both models showed the ability to follow seasonal rainfall patterns in some evaluation blocks; however, SARIMA demonstrated more stable and consistent performance across the temporal blocks than AMBP. The results indicate that linear statistical models remain effective in modeling rainfall with strong seasonal patterns and have the potential to support water resource planning, hydrometeorological disaster mitigation, and information-based agricultural sector management.]]>
Copyrights © 2026
