Daily Rainfall Prediction and Intensity Classification in Sleman Regency Using Random Forest Regressor

Abstract

Daily rainfall forecasting is essential for water resource management, agricultural planning, and hydrometeorological disaster mitigation, especially in tropical regions such as Indonesia that experience high climate variability. This study aims to develop a daily rainfall prediction model for Sleman Regency using a Random Forest Regressor (RFR) involving 11 input features, consisting of eight surface meteorological variables and three lag variables based on daily data from 2015 to 2025. This model is designed to capture non-linear relationships and temporal dependencies on rainfall patterns. In the test data, the model achieved a Mean Absolute Error (MAE) of 7.82 mm, a Root Mean Squared Error (RMSE) of 15.34 mm, and an R² of 0.396, which explains about 39.6% of the variation in daily rainfall. To improve the interpretation of the prediction results, the output of this model is classified into the category of rain intensity in accordance with BMKG standards. The results showed that the Random Forest Regressor was effective for modeling general rainfall trends and temporal patterns, but difficulties in predicting extreme rainfall events, mainly due to class imbalances and ensemble averages. This study provides a practical framework for combining non-linear regression with operational rainfall intensity classification using long-term daily data, advancing computational approaches for environmental intelligence in Informatics.