This study aims to compare the performance of automatic and manual rain gauges in the Central Highlands of Central Java using a machine learning approach based on Extreme Gradient Boosting (XGBoost) and Random Forest Regression (RFR) algorithms. Daily rainfall data were collected from five regencies—Banyumas, Banjarnegara, Wonosobo, Temanggung, and Pemalang—between 2021 and 2024. Preprocessing involved merging data from two types of instruments (Automatic Rain Gauge/AWS and manual ombrometer), correcting anomalies, and standardizing date-time formats. The models were developed using feature engineering techniques, including multi-lag and moving averages, and evaluated using MAE, RMSE, and R-squared (R²) metrics. The results show that the XGBoost model with automatic data achieved the best performance, with a Mean Absolute Error (MAE) of 17.3632 mm, Root Mean Squared Error (RMSE) of 27.0282 mm, and R² of 0.5050. In comparison, the Random Forest model with automatically generated data produced an MAE of 16.6307 mm, an RMSE of 28.5286 mm, and an R² of 0.4485. Models with manual data showed lower performance, with R² values below 0.30. These findings indicate that automatic measurement data are more stable and effective for building predictive rainfall models using machine learning. This supports the use of automatic instruments as the primary data source in rainfall forecasting and hydrometeorological disaster mitigation systems.