Classification of Potential Tsunami Disaster Due to Earthquakes in Indonesia Based on Machine Learning

Abstract

Earthquakes and tsunamis pose significant threats to Indonesia due to its unique geological positioning at the convergence of four tectonic plates. This study focuses on classifying the potential occurrence of tsunami disasters following earthquakes using various data mining methods, including k-Nearest Neighbor (kNN), Naïve Bayes, Decision Tree and Ensemble Method, and Linear Regression. The research employs a qualitative approach to systematically understand and describe the context of natural disasters, utilizing both primary and secondary data collection techniques. Performance evaluation metrics such as Area Under the Curve (AUC), Classification Accuracy (CA), F1 Score, Precision, and Recall are utilized to assess the effectiveness of each method in predicting potential tsunami events. The findings reveal that the kNN method exhibits the highest performance, with an AUC of 94.4% and a precision of 82.8%, indicating robust predictive capabilities. However, misclassifications were observed, emphasizing the need for further refinement. Naïve Bayes also shows promising results with an AUC of 84.5% and precision of 78.6%. Decision Tree and Ensemble Method models, such as Random Forest and AdaBoost, demonstrate reasonable performance, with Random Forest achieving the highest AUC of 71.9%. Linear Regression is employed to explore the correlation between earthquake attributes and tsunami occurrence, revealing a weak relationship. Further research integrating advanced modeling approaches and additional earthquake attributes is recommended to enhance the predictive capabilities of tsunami risk assessment models. The study underscores the importance of employing diverse machine learning techniques and evaluating their performance metrics to refine the accuracy of tsunami prediction models, ultimately contributing to practical disaster preparedness and mitigation strategies.