Advanced Machine Learning Techniques for Seismic Anomaly Detection in Indonesia: A Comparative Study of Lof, Isolation Forest, and One-Class SVM

Abstract

This study presents a comprehensive comparison of three machine learning algorithms for anomaly detection within seismic data, focusing on the unique geographical and geological context of Indonesia, a region prone to frequent seismic events. Local Outlier Factor (LOF), Isolation Forest, and One-Class SVM were assessed using a meticulously curated dataset from the Indonesian Meteorology, Climatology, and Geophysical Agency, standardized to ensure consistent feature scale. Our analysis encompassed both statistical metrics and visualizations to evaluate the performance of each algorithm. The One-Class SVM emerged as the most effective method, achieving the highest silhouette score, indicative of its superior cluster formation and clear distinction between inliers and outliers. The Isolation Forest also demonstrated strong performance with a favorable silhouette score and Davies-Bouldin index, suggesting effective anomaly isolation capabilities. In contrast, the LOF algorithm showed less precision, as indicated by lower silhouette scores and a higher Davies-Bouldin index, suggesting potential challenges in distinguishing between normal and anomalous seismic patterns. Statistical validation using the Kruskal-Wallis H-test confirmed significant differences in the anomaly score distributions of the three algorithms, with a p-value of 0.0. Visualizations through PCA and t-SNE reinforced the quantitative findings, displaying a clear demarcation of anomalies by the One-Class SVM and Isolation Forest, unlike the LOF.The findings underscore the importance of selecting appropriate anomaly detection methods for seismic data analysis, highlighting the robustness of One-Class SVM and Isolation Forest for such applications. The implications of this research are profound for seismic risk management, providing insights that enhance the accuracy and reliability of earthquake prediction systems, which is vital for regions with high seismic activity such as Indonesia.