<![CDATA[Electricity is a critical resource that supports various sectors in Indonesia, especially during extreme weather. Outages have become serious for operational risks during extreme weather. This study proposes a machine learning-based early warning system to predict electricity outages caused by extreme weather. Historical weather and outage data were combined using spatial alignment. Key innovation of this study involved geospatial feature enrichment via HDBSCAN, Yeo-Johnson transformation, robust scaling, and class resampling using SMOTE, ADASYN, and SMOTE-ENN. Four ensemble classification models (Random Forest, XGBoost, AdaBoost, and LightGBM) were evaluated. LightGBM with SMOTE yielded the highest recall (0.99) and the fewest false negatives. These findings suggest a solution for a proactive early warning system risk mitigation in electricity under extreme weather conditions.]]>
Copyrights © 2025
