<![CDATA[Rainfall is one of the important meteorological elements in various sectors, such as agriculture, water resource management, and hydrometeorological disaster mitigation. This study aims to compare the performance of Seasonal Autoregressive Integrated Moving Average with Exogenous Variables (SARIMAX), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM) models in monthly rainfall forecasting in Makassar City. The data used were monthly meteorological data from January 2000 to December 2024 obtained from NASA POWER with a total of 300 observations. The variables used include rainfall, temperature, humidity, wind speed, pressure, and solar radiation. The research stages consisted of data preprocessing, exploratory data analysis, stationarity testing using the Augmented Dickey-Fuller (ADF) test, SARIMAX, XGBoost, and LSTM modeling, and model evaluation using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R^2). The results showed that the XGBoost model achieved the best performance with an RMSE value of 2.5441 and an R^2 value of 0.8063, while the SARIMAX model produced the lowest MAPE value of 39.2614%. Meanwhile, the LSTM model showed less optimal performance with an RMSE value of 5.2125 and an R^2 value of 0.1646. The results indicate that the boosting-based machine learning approach is more effective in handling nonlinear relationships in monthly rainfall data compared to classical statistical and deep learning models on limited datasets.]]>
Copyrights © 2026
