Perbandingan Model Decision Tree untuk Prediksi Intensitas Hujan di Juanda, Sidoarjo, Jawa Timur

Abstract

Rainfall intensity is one of the key parameters in climate dynamics and is strongly associated with the increasing occurrence of hydrometeorological disasters. This study aims to evaluate and compare the performance of three decision tree–based machine learning models in predicting rainfall intensity in the Juanda region, Sidoarjo, East Java. The data used consist of daily weather observations from the Juanda Class I Meteorological Station during the period 2018–2022, covering 13 meteorological variables. The models compared include the Extra Trees Classifier, Random Forest Classifier, and XGBoost Classifier, with performance evaluation based on accuracy, precision, recall, F1 score, and Area Under the Curve (AUC). The results indicate that the Extra Trees Classifier demonstrates the best performance, with an accuracy of 0.8123; precision of 0.8151; recall of 0.8123; AUC of 0.9158; and F1 score of 0.8126. Relative humidity (Rh) was identified as the most influential variable in predicting rainfall intensity across all three models. These findings provide further insights into the relationship between daily weather parameters and rainfall intensity and contribute to the development of more accurate predictive systems as a basis for hydrometeorological disaster mitigation in the study area.