Perbandingan Algoritma Machine Learning (Logistic Regression, SVM, KNN, Decision Tree, Random Forest, dan Gradient Boosting) dalam Prediksi Hujan Harian di Provinsi Lampung

Abstract

In this era, rainfall prediction has become important for meteorological analysis, natural resource management, and hydrometeorological disaster mitigation. The development of rainfall prediction is aligned with modern predictive models, such as popular Machine Learning (ML) algorithms. Given these two phenomena, this study aims to evaluate six ML algorithms: Logistic Regression (LR), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), and Gradient Boosting Machine (GBM). Rainfall prediction combines data from three sources, namely ERA5, NASA POWER, and BMKG Lampung. The data was taken from three BMKG Lampung observation stations (Bandar Lampung, Pesawaran, and Kotabumi) for the period 2020-2025. Performance evaluation was conducted by assessing accuracy, precision, recall, F1-Score, confusion matrix, ROC curve, and AUC. The results of the study show that Random Forest achieves the best performance across all aspects, with a UAC value of 0.715. AUC value indicates that RF has stable predictions for very complex rainfall cases. The Decision Tree model showed low accuracy (0.66) and AUC values of 0.579. Ensemble-based multiple trees are more effective for predictive performance. These results highlight the importance of integrating various meteorological datasets from multiple sources using advanced ensemble learning methods to improve rainfall prediction accuracy and support climate-related decision-making at the regional level.