Analyzing the Performance of Machine Learning Architectures in Predicting Padang's Precipitation

Rahmat Hidayat (University of People)

Inferensi
Vol 9 No 1 (2026)

This study evaluates the effectiveness of a machine learning approach for delivering precise daily rainfall forecasts in Padang City. The study aims to determine the most effective predictive model to support local decision-making and urban development, taking into account the considerable variability of precipitation patterns in the area. The methodology involves a comparative analysis of three prominent machine learning algorithms: Logistic Regression, Random Forest, and Extreme Gradient Boosting (XGBoost). Each model was meticulously evaluated using a comprehensive set of criteria, including accuracy, precision, recall, F1 score, and Area Under the Curve (AUC). The experimental findings indicate that all three models can forecast daily precipitation with considerable accuracy. The Random Forest model demonstrated superior performance within the group, achieving a peak prediction accuracy of 85%. The statistics demonstrate that the Random Forest model is the most dependable approach for forecasting precipitation events in Padang City. This model is highly recommended for integration into early warning systems and activity planning frameworks to mitigate the impacts of unpredictable weather in urban environments.