<![CDATA[Earthquakes are one of the most destructive natural disasters, particularly in Indonesia, which is located at the convergence of three active tectonic plates. Conventional early warning systems generally rely on real-time vibration detection but lack the capability to provide comprehensive predictions about the potential severity of an earthquake. This study aims to address these limitations by applying data mining techniques and machine learning algorithms to classify earthquake alert levels based on seismic parameters, including magnitude, depth, Community Determined Intensity (CDI), Modified Mercalli Intensity (MMI), and significance (Sig). A dataset of 1,300 earthquake records was obtained and processed using the Knowledge Discovery in Database (KDD) methodology, which includes data selection, preprocessing, transformation, modeling, and evaluation. Five classification algorithms were compared: Decision Tree, Random Forest, Naïve Bayes, K-Nearest Neighbor (KNN), and Neural Network. Model performance was evaluated using confusion matrix metrics such as accuracy, precision, recall, and F1-score. The results indicate that Random Forest achieved the highest performance with an accuracy of 88.52% and macro recall of 88.90%, outperforming other algorithms. Decision Tree ranked second with balanced performance, while KNN and Neural Network achieved moderate results. Naïve Bayes performed the weakest. Overall, Random Forest is the most reliable algorithm for supporting earthquake early warning systems.]]>
Copyrights © 2025
